skbuff.h 97 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914915916917918919920921922923924925926927928929930931932933934935936937938939940941942943944945946947948949950951952953954955956957958959960961962963964965966967968969970971972973974975976977978979980981982983984985986987988989990991992993994995996997998999100010011002100310041005100610071008100910101011101210131014101510161017101810191020102110221023102410251026102710281029103010311032103310341035103610371038103910401041104210431044104510461047104810491050105110521053105410551056105710581059106010611062106310641065106610671068106910701071107210731074107510761077107810791080108110821083108410851086108710881089109010911092109310941095109610971098109911001101110211031104110511061107110811091110111111121113111411151116111711181119112011211122112311241125112611271128112911301131113211331134113511361137113811391140114111421143114411451146114711481149115011511152115311541155115611571158115911601161116211631164116511661167116811691170117111721173117411751176117711781179118011811182118311841185118611871188118911901191119211931194119511961197119811991200120112021203120412051206120712081209121012111212121312141215121612171218121912201221122212231224122512261227122812291230123112321233123412351236123712381239124012411242124312441245124612471248124912501251125212531254125512561257125812591260126112621263126412651266126712681269127012711272127312741275127612771278127912801281128212831284128512861287128812891290129112921293129412951296129712981299130013011302130313041305130613071308130913101311131213131314131513161317131813191320132113221323132413251326132713281329133013311332133313341335133613371338133913401341134213431344134513461347134813491350135113521353135413551356135713581359136013611362136313641365136613671368136913701371137213731374137513761377137813791380138113821383138413851386138713881389139013911392139313941395139613971398139914001401140214031404140514061407140814091410141114121413141414151416141714181419142014211422142314241425142614271428142914301431143214331434143514361437143814391440144114421443144414451446144714481449145014511452145314541455145614571458145914601461146214631464146514661467146814691470147114721473147414751476147714781479148014811482148314841485148614871488148914901491149214931494149514961497149814991500150115021503150415051506150715081509151015111512151315141515151615171518151915201521152215231524152515261527152815291530153115321533153415351536153715381539154015411542154315441545154615471548154915501551155215531554155515561557155815591560156115621563156415651566156715681569157015711572157315741575157615771578157915801581158215831584158515861587158815891590159115921593159415951596159715981599160016011602160316041605160616071608160916101611161216131614161516161617161816191620162116221623162416251626162716281629163016311632163316341635163616371638163916401641164216431644164516461647164816491650165116521653165416551656165716581659166016611662166316641665166616671668166916701671167216731674167516761677167816791680168116821683168416851686168716881689169016911692169316941695169616971698169917001701170217031704170517061707170817091710171117121713171417151716171717181719172017211722172317241725172617271728172917301731173217331734173517361737173817391740174117421743174417451746174717481749175017511752175317541755175617571758175917601761176217631764176517661767176817691770177117721773177417751776177717781779178017811782178317841785178617871788178917901791179217931794179517961797179817991800180118021803180418051806180718081809181018111812181318141815181618171818181918201821182218231824182518261827182818291830183118321833183418351836183718381839184018411842184318441845184618471848184918501851185218531854185518561857185818591860186118621863186418651866186718681869187018711872187318741875187618771878187918801881188218831884188518861887188818891890189118921893189418951896189718981899190019011902190319041905190619071908190919101911191219131914191519161917191819191920192119221923192419251926192719281929193019311932193319341935193619371938193919401941194219431944194519461947194819491950195119521953195419551956195719581959196019611962196319641965196619671968196919701971197219731974197519761977197819791980198119821983198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007200820092010201120122013201420152016201720182019202020212022202320242025202620272028202920302031203220332034203520362037203820392040204120422043204420452046204720482049205020512052205320542055205620572058205920602061206220632064206520662067206820692070207120722073207420752076207720782079208020812082208320842085208620872088208920902091209220932094209520962097209820992100210121022103210421052106210721082109211021112112211321142115211621172118211921202121212221232124212521262127212821292130213121322133213421352136213721382139214021412142214321442145214621472148214921502151215221532154215521562157215821592160216121622163216421652166216721682169217021712172217321742175217621772178217921802181218221832184218521862187218821892190219121922193219421952196219721982199220022012202220322042205220622072208220922102211221222132214221522162217221822192220222122222223222422252226222722282229223022312232223322342235223622372238223922402241224222432244224522462247224822492250225122522253225422552256225722582259226022612262226322642265226622672268226922702271227222732274227522762277227822792280228122822283228422852286228722882289229022912292229322942295229622972298229923002301230223032304230523062307230823092310231123122313231423152316231723182319232023212322232323242325232623272328232923302331233223332334233523362337233823392340234123422343234423452346234723482349235023512352235323542355235623572358235923602361236223632364236523662367236823692370237123722373237423752376237723782379238023812382238323842385238623872388238923902391239223932394239523962397239823992400240124022403240424052406240724082409241024112412241324142415241624172418241924202421242224232424242524262427242824292430243124322433243424352436243724382439244024412442244324442445244624472448244924502451245224532454245524562457245824592460246124622463246424652466246724682469247024712472247324742475247624772478247924802481248224832484248524862487248824892490249124922493249424952496249724982499250025012502250325042505250625072508250925102511251225132514251525162517251825192520252125222523252425252526252725282529253025312532253325342535253625372538253925402541254225432544254525462547254825492550255125522553255425552556255725582559256025612562256325642565256625672568256925702571257225732574257525762577257825792580258125822583258425852586258725882589259025912592259325942595259625972598259926002601260226032604260526062607260826092610261126122613261426152616261726182619262026212622262326242625262626272628262926302631263226332634263526362637263826392640264126422643264426452646264726482649265026512652265326542655265626572658265926602661266226632664266526662667266826692670267126722673267426752676267726782679268026812682268326842685268626872688268926902691269226932694269526962697269826992700270127022703270427052706270727082709271027112712271327142715271627172718271927202721272227232724272527262727272827292730273127322733273427352736273727382739274027412742274327442745274627472748274927502751275227532754275527562757275827592760276127622763276427652766276727682769277027712772277327742775277627772778277927802781278227832784278527862787278827892790279127922793279427952796279727982799280028012802280328042805280628072808280928102811281228132814281528162817281828192820282128222823282428252826282728282829283028312832283328342835283628372838283928402841284228432844284528462847284828492850285128522853285428552856285728582859286028612862286328642865286628672868286928702871287228732874287528762877287828792880288128822883288428852886288728882889289028912892289328942895289628972898289929002901290229032904290529062907290829092910291129122913291429152916291729182919292029212922292329242925292629272928292929302931293229332934293529362937293829392940294129422943294429452946294729482949295029512952295329542955295629572958295929602961296229632964296529662967296829692970297129722973297429752976297729782979298029812982298329842985298629872988298929902991299229932994299529962997299829993000300130023003300430053006300730083009301030113012301330143015301630173018301930203021302230233024302530263027302830293030303130323033303430353036303730383039304030413042304330443045304630473048304930503051305230533054305530563057305830593060306130623063306430653066306730683069307030713072307330743075307630773078307930803081308230833084308530863087308830893090309130923093309430953096309730983099310031013102310331043105310631073108310931103111311231133114311531163117311831193120312131223123312431253126312731283129313031313132313331343135313631373138313931403141314231433144314531463147314831493150315131523153315431553156315731583159316031613162316331643165316631673168316931703171317231733174317531763177317831793180318131823183318431853186318731883189319031913192319331943195319631973198319932003201320232033204320532063207320832093210321132123213321432153216321732183219322032213222322332243225322632273228322932303231323232333234323532363237323832393240324132423243324432453246324732483249325032513252325332543255325632573258325932603261326232633264326532663267326832693270327132723273327432753276327732783279328032813282328332843285328632873288328932903291329232933294329532963297329832993300330133023303330433053306330733083309331033113312331333143315331633173318331933203321332233233324332533263327332833293330333133323333333433353336333733383339334033413342334333443345334633473348334933503351335233533354335533563357335833593360336133623363336433653366336733683369337033713372337333743375337633773378337933803381338233833384338533863387338833893390339133923393339433953396339733983399340034013402340334043405340634073408340934103411341234133414341534163417341834193420342134223423342434253426342734283429343034313432343334343435343634373438343934403441344234433444344534463447344834493450345134523453345434553456
  1. /*
  2. * Definitions for the 'struct sk_buff' memory handlers.
  3. *
  4. * Authors:
  5. * Alan Cox, <gw4pts@gw4pts.ampr.org>
  6. * Florian La Roche, <rzsfl@rz.uni-sb.de>
  7. *
  8. * This program is free software; you can redistribute it and/or
  9. * modify it under the terms of the GNU General Public License
  10. * as published by the Free Software Foundation; either version
  11. * 2 of the License, or (at your option) any later version.
  12. */
  13. #ifndef _LINUX_SKBUFF_H
  14. #define _LINUX_SKBUFF_H
  15. #include <linux/kernel.h>
  16. #include <linux/kmemcheck.h>
  17. #include <linux/compiler.h>
  18. #include <linux/time.h>
  19. #include <linux/bug.h>
  20. #include <linux/cache.h>
  21. #include <linux/rbtree.h>
  22. #include <linux/socket.h>
  23. #include <linux/atomic.h>
  24. #include <asm/types.h>
  25. #include <linux/spinlock.h>
  26. #include <linux/net.h>
  27. #include <linux/textsearch.h>
  28. #include <net/checksum.h>
  29. #include <linux/rcupdate.h>
  30. #include <linux/hrtimer.h>
  31. #include <linux/dma-mapping.h>
  32. #include <linux/netdev_features.h>
  33. #include <linux/sched.h>
  34. #include <net/flow_keys.h>
  35. /* A. Checksumming of received packets by device.
  36. *
  37. * CHECKSUM_NONE:
  38. *
  39. * Device failed to checksum this packet e.g. due to lack of capabilities.
  40. * The packet contains full (though not verified) checksum in packet but
  41. * not in skb->csum. Thus, skb->csum is undefined in this case.
  42. *
  43. * CHECKSUM_UNNECESSARY:
  44. *
  45. * The hardware you're dealing with doesn't calculate the full checksum
  46. * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
  47. * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
  48. * if their checksums are okay. skb->csum is still undefined in this case
  49. * though. It is a bad option, but, unfortunately, nowadays most vendors do
  50. * this. Apparently with the secret goal to sell you new devices, when you
  51. * will add new protocol to your host, f.e. IPv6 8)
  52. *
  53. * CHECKSUM_UNNECESSARY is applicable to following protocols:
  54. * TCP: IPv6 and IPv4.
  55. * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
  56. * zero UDP checksum for either IPv4 or IPv6, the networking stack
  57. * may perform further validation in this case.
  58. * GRE: only if the checksum is present in the header.
  59. * SCTP: indicates the CRC in SCTP header has been validated.
  60. *
  61. * skb->csum_level indicates the number of consecutive checksums found in
  62. * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
  63. * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
  64. * and a device is able to verify the checksums for UDP (possibly zero),
  65. * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
  66. * two. If the device were only able to verify the UDP checksum and not
  67. * GRE, either because it doesn't support GRE checksum of because GRE
  68. * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
  69. * not considered in this case).
  70. *
  71. * CHECKSUM_COMPLETE:
  72. *
  73. * This is the most generic way. The device supplied checksum of the _whole_
  74. * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
  75. * hardware doesn't need to parse L3/L4 headers to implement this.
  76. *
  77. * Note: Even if device supports only some protocols, but is able to produce
  78. * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
  79. *
  80. * CHECKSUM_PARTIAL:
  81. *
  82. * A checksum is set up to be offloaded to a device as described in the
  83. * output description for CHECKSUM_PARTIAL. This may occur on a packet
  84. * received directly from another Linux OS, e.g., a virtualized Linux kernel
  85. * on the same host, or it may be set in the input path in GRO or remote
  86. * checksum offload. For the purposes of checksum verification, the checksum
  87. * referred to by skb->csum_start + skb->csum_offset and any preceding
  88. * checksums in the packet are considered verified. Any checksums in the
  89. * packet that are after the checksum being offloaded are not considered to
  90. * be verified.
  91. *
  92. * B. Checksumming on output.
  93. *
  94. * CHECKSUM_NONE:
  95. *
  96. * The skb was already checksummed by the protocol, or a checksum is not
  97. * required.
  98. *
  99. * CHECKSUM_PARTIAL:
  100. *
  101. * The device is required to checksum the packet as seen by hard_start_xmit()
  102. * from skb->csum_start up to the end, and to record/write the checksum at
  103. * offset skb->csum_start + skb->csum_offset.
  104. *
  105. * The device must show its capabilities in dev->features, set up at device
  106. * setup time, e.g. netdev_features.h:
  107. *
  108. * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
  109. * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
  110. * IPv4. Sigh. Vendors like this way for an unknown reason.
  111. * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
  112. * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
  113. * NETIF_F_... - Well, you get the picture.
  114. *
  115. * CHECKSUM_UNNECESSARY:
  116. *
  117. * Normally, the device will do per protocol specific checksumming. Protocol
  118. * implementations that do not want the NIC to perform the checksum
  119. * calculation should use this flag in their outgoing skbs.
  120. *
  121. * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
  122. * offload. Correspondingly, the FCoE protocol driver
  123. * stack should use CHECKSUM_UNNECESSARY.
  124. *
  125. * Any questions? No questions, good. --ANK
  126. */
  127. /* Don't change this without changing skb_csum_unnecessary! */
  128. #define CHECKSUM_NONE 0
  129. #define CHECKSUM_UNNECESSARY 1
  130. #define CHECKSUM_COMPLETE 2
  131. #define CHECKSUM_PARTIAL 3
  132. /* Maximum value in skb->csum_level */
  133. #define SKB_MAX_CSUM_LEVEL 3
  134. #define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
  135. #define SKB_WITH_OVERHEAD(X) \
  136. ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  137. #define SKB_MAX_ORDER(X, ORDER) \
  138. SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
  139. #define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
  140. #define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
  141. /* return minimum truesize of one skb containing X bytes of data */
  142. #define SKB_TRUESIZE(X) ((X) + \
  143. SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
  144. SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
  145. struct net_device;
  146. struct scatterlist;
  147. struct pipe_inode_info;
  148. struct iov_iter;
  149. struct napi_struct;
  150. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  151. struct nf_conntrack {
  152. atomic_t use;
  153. };
  154. #endif
  155. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  156. struct nf_bridge_info {
  157. atomic_t use;
  158. enum {
  159. BRNF_PROTO_UNCHANGED,
  160. BRNF_PROTO_8021Q,
  161. BRNF_PROTO_PPPOE
  162. } orig_proto;
  163. bool pkt_otherhost;
  164. unsigned int mask;
  165. struct net_device *physindev;
  166. struct net_device *physoutdev;
  167. char neigh_header[8];
  168. };
  169. #endif
  170. struct sk_buff_head {
  171. /* These two members must be first. */
  172. struct sk_buff *next;
  173. struct sk_buff *prev;
  174. __u32 qlen;
  175. spinlock_t lock;
  176. };
  177. struct sk_buff;
  178. /* To allow 64K frame to be packed as single skb without frag_list we
  179. * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
  180. * buffers which do not start on a page boundary.
  181. *
  182. * Since GRO uses frags we allocate at least 16 regardless of page
  183. * size.
  184. */
  185. #if (65536/PAGE_SIZE + 1) < 16
  186. #define MAX_SKB_FRAGS 16UL
  187. #else
  188. #define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
  189. #endif
  190. typedef struct skb_frag_struct skb_frag_t;
  191. struct skb_frag_struct {
  192. struct {
  193. struct page *p;
  194. } page;
  195. #if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
  196. __u32 page_offset;
  197. __u32 size;
  198. #else
  199. __u16 page_offset;
  200. __u16 size;
  201. #endif
  202. };
  203. static inline unsigned int skb_frag_size(const skb_frag_t *frag)
  204. {
  205. return frag->size;
  206. }
  207. static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
  208. {
  209. frag->size = size;
  210. }
  211. static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
  212. {
  213. frag->size += delta;
  214. }
  215. static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
  216. {
  217. frag->size -= delta;
  218. }
  219. #define HAVE_HW_TIME_STAMP
  220. /**
  221. * struct skb_shared_hwtstamps - hardware time stamps
  222. * @hwtstamp: hardware time stamp transformed into duration
  223. * since arbitrary point in time
  224. *
  225. * Software time stamps generated by ktime_get_real() are stored in
  226. * skb->tstamp.
  227. *
  228. * hwtstamps can only be compared against other hwtstamps from
  229. * the same device.
  230. *
  231. * This structure is attached to packets as part of the
  232. * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
  233. */
  234. struct skb_shared_hwtstamps {
  235. ktime_t hwtstamp;
  236. };
  237. /* Definitions for tx_flags in struct skb_shared_info */
  238. enum {
  239. /* generate hardware time stamp */
  240. SKBTX_HW_TSTAMP = 1 << 0,
  241. /* generate software time stamp when queueing packet to NIC */
  242. SKBTX_SW_TSTAMP = 1 << 1,
  243. /* device driver is going to provide hardware time stamp */
  244. SKBTX_IN_PROGRESS = 1 << 2,
  245. /* device driver supports TX zero-copy buffers */
  246. SKBTX_DEV_ZEROCOPY = 1 << 3,
  247. /* generate wifi status information (where possible) */
  248. SKBTX_WIFI_STATUS = 1 << 4,
  249. /* This indicates at least one fragment might be overwritten
  250. * (as in vmsplice(), sendfile() ...)
  251. * If we need to compute a TX checksum, we'll need to copy
  252. * all frags to avoid possible bad checksum
  253. */
  254. SKBTX_SHARED_FRAG = 1 << 5,
  255. /* generate software time stamp when entering packet scheduling */
  256. SKBTX_SCHED_TSTAMP = 1 << 6,
  257. /* generate software timestamp on peer data acknowledgment */
  258. SKBTX_ACK_TSTAMP = 1 << 7,
  259. };
  260. #define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
  261. SKBTX_SCHED_TSTAMP | \
  262. SKBTX_ACK_TSTAMP)
  263. #define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
  264. /*
  265. * The callback notifies userspace to release buffers when skb DMA is done in
  266. * lower device, the skb last reference should be 0 when calling this.
  267. * The zerocopy_success argument is true if zero copy transmit occurred,
  268. * false on data copy or out of memory error caused by data copy attempt.
  269. * The ctx field is used to track device context.
  270. * The desc field is used to track userspace buffer index.
  271. */
  272. struct ubuf_info {
  273. void (*callback)(struct ubuf_info *, bool zerocopy_success);
  274. void *ctx;
  275. unsigned long desc;
  276. };
  277. /* This data is invariant across clones and lives at
  278. * the end of the header data, ie. at skb->end.
  279. */
  280. struct skb_shared_info {
  281. unsigned char nr_frags;
  282. __u8 tx_flags;
  283. unsigned short gso_size;
  284. /* Warning: this field is not always filled in (UFO)! */
  285. unsigned short gso_segs;
  286. unsigned short gso_type;
  287. struct sk_buff *frag_list;
  288. struct skb_shared_hwtstamps hwtstamps;
  289. u32 tskey;
  290. __be32 ip6_frag_id;
  291. /*
  292. * Warning : all fields before dataref are cleared in __alloc_skb()
  293. */
  294. atomic_t dataref;
  295. /* Intermediate layers must ensure that destructor_arg
  296. * remains valid until skb destructor */
  297. void * destructor_arg;
  298. /* must be last field, see pskb_expand_head() */
  299. skb_frag_t frags[MAX_SKB_FRAGS];
  300. };
  301. /* We divide dataref into two halves. The higher 16 bits hold references
  302. * to the payload part of skb->data. The lower 16 bits hold references to
  303. * the entire skb->data. A clone of a headerless skb holds the length of
  304. * the header in skb->hdr_len.
  305. *
  306. * All users must obey the rule that the skb->data reference count must be
  307. * greater than or equal to the payload reference count.
  308. *
  309. * Holding a reference to the payload part means that the user does not
  310. * care about modifications to the header part of skb->data.
  311. */
  312. #define SKB_DATAREF_SHIFT 16
  313. #define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
  314. enum {
  315. SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
  316. SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
  317. SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
  318. };
  319. enum {
  320. SKB_GSO_TCPV4 = 1 << 0,
  321. SKB_GSO_UDP = 1 << 1,
  322. /* This indicates the skb is from an untrusted source. */
  323. SKB_GSO_DODGY = 1 << 2,
  324. /* This indicates the tcp segment has CWR set. */
  325. SKB_GSO_TCP_ECN = 1 << 3,
  326. SKB_GSO_TCPV6 = 1 << 4,
  327. SKB_GSO_FCOE = 1 << 5,
  328. SKB_GSO_GRE = 1 << 6,
  329. SKB_GSO_GRE_CSUM = 1 << 7,
  330. SKB_GSO_IPIP = 1 << 8,
  331. SKB_GSO_SIT = 1 << 9,
  332. SKB_GSO_UDP_TUNNEL = 1 << 10,
  333. SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
  334. SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
  335. };
  336. #if BITS_PER_LONG > 32
  337. #define NET_SKBUFF_DATA_USES_OFFSET 1
  338. #endif
  339. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  340. typedef unsigned int sk_buff_data_t;
  341. #else
  342. typedef unsigned char *sk_buff_data_t;
  343. #endif
  344. /**
  345. * struct skb_mstamp - multi resolution time stamps
  346. * @stamp_us: timestamp in us resolution
  347. * @stamp_jiffies: timestamp in jiffies
  348. */
  349. struct skb_mstamp {
  350. union {
  351. u64 v64;
  352. struct {
  353. u32 stamp_us;
  354. u32 stamp_jiffies;
  355. };
  356. };
  357. };
  358. /**
  359. * skb_mstamp_get - get current timestamp
  360. * @cl: place to store timestamps
  361. */
  362. static inline void skb_mstamp_get(struct skb_mstamp *cl)
  363. {
  364. u64 val = local_clock();
  365. do_div(val, NSEC_PER_USEC);
  366. cl->stamp_us = (u32)val;
  367. cl->stamp_jiffies = (u32)jiffies;
  368. }
  369. /**
  370. * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
  371. * @t1: pointer to newest sample
  372. * @t0: pointer to oldest sample
  373. */
  374. static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
  375. const struct skb_mstamp *t0)
  376. {
  377. s32 delta_us = t1->stamp_us - t0->stamp_us;
  378. u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
  379. /* If delta_us is negative, this might be because interval is too big,
  380. * or local_clock() drift is too big : fallback using jiffies.
  381. */
  382. if (delta_us <= 0 ||
  383. delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
  384. delta_us = jiffies_to_usecs(delta_jiffies);
  385. return delta_us;
  386. }
  387. /**
  388. * struct sk_buff - socket buffer
  389. * @next: Next buffer in list
  390. * @prev: Previous buffer in list
  391. * @tstamp: Time we arrived/left
  392. * @rbnode: RB tree node, alternative to next/prev for netem/tcp
  393. * @sk: Socket we are owned by
  394. * @dev: Device we arrived on/are leaving by
  395. * @cb: Control buffer. Free for use by every layer. Put private vars here
  396. * @_skb_refdst: destination entry (with norefcount bit)
  397. * @sp: the security path, used for xfrm
  398. * @len: Length of actual data
  399. * @data_len: Data length
  400. * @mac_len: Length of link layer header
  401. * @hdr_len: writable header length of cloned skb
  402. * @csum: Checksum (must include start/offset pair)
  403. * @csum_start: Offset from skb->head where checksumming should start
  404. * @csum_offset: Offset from csum_start where checksum should be stored
  405. * @priority: Packet queueing priority
  406. * @ignore_df: allow local fragmentation
  407. * @cloned: Head may be cloned (check refcnt to be sure)
  408. * @ip_summed: Driver fed us an IP checksum
  409. * @nohdr: Payload reference only, must not modify header
  410. * @nfctinfo: Relationship of this skb to the connection
  411. * @pkt_type: Packet class
  412. * @fclone: skbuff clone status
  413. * @ipvs_property: skbuff is owned by ipvs
  414. * @peeked: this packet has been seen already, so stats have been
  415. * done for it, don't do them again
  416. * @nf_trace: netfilter packet trace flag
  417. * @protocol: Packet protocol from driver
  418. * @destructor: Destruct function
  419. * @nfct: Associated connection, if any
  420. * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
  421. * @skb_iif: ifindex of device we arrived on
  422. * @tc_index: Traffic control index
  423. * @tc_verd: traffic control verdict
  424. * @hash: the packet hash
  425. * @queue_mapping: Queue mapping for multiqueue devices
  426. * @xmit_more: More SKBs are pending for this queue
  427. * @ndisc_nodetype: router type (from link layer)
  428. * @ooo_okay: allow the mapping of a socket to a queue to be changed
  429. * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
  430. * ports.
  431. * @sw_hash: indicates hash was computed in software stack
  432. * @wifi_acked_valid: wifi_acked was set
  433. * @wifi_acked: whether frame was acked on wifi or not
  434. * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
  435. * @napi_id: id of the NAPI struct this skb came from
  436. * @secmark: security marking
  437. * @mark: Generic packet mark
  438. * @vlan_proto: vlan encapsulation protocol
  439. * @vlan_tci: vlan tag control information
  440. * @inner_protocol: Protocol (encapsulation)
  441. * @inner_transport_header: Inner transport layer header (encapsulation)
  442. * @inner_network_header: Network layer header (encapsulation)
  443. * @inner_mac_header: Link layer header (encapsulation)
  444. * @transport_header: Transport layer header
  445. * @network_header: Network layer header
  446. * @mac_header: Link layer header
  447. * @tail: Tail pointer
  448. * @end: End pointer
  449. * @head: Head of buffer
  450. * @data: Data head pointer
  451. * @truesize: Buffer size
  452. * @users: User count - see {datagram,tcp}.c
  453. */
  454. struct sk_buff {
  455. union {
  456. struct {
  457. /* These two members must be first. */
  458. struct sk_buff *next;
  459. struct sk_buff *prev;
  460. union {
  461. ktime_t tstamp;
  462. struct skb_mstamp skb_mstamp;
  463. };
  464. };
  465. struct rb_node rbnode; /* used in netem & tcp stack */
  466. };
  467. struct sock *sk;
  468. struct net_device *dev;
  469. /*
  470. * This is the control buffer. It is free to use for every
  471. * layer. Please put your private variables there. If you
  472. * want to keep them across layers you have to do a skb_clone()
  473. * first. This is owned by whoever has the skb queued ATM.
  474. */
  475. char cb[48] __aligned(8);
  476. unsigned long _skb_refdst;
  477. void (*destructor)(struct sk_buff *skb);
  478. #ifdef CONFIG_XFRM
  479. struct sec_path *sp;
  480. #endif
  481. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  482. struct nf_conntrack *nfct;
  483. #endif
  484. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  485. struct nf_bridge_info *nf_bridge;
  486. #endif
  487. unsigned int len,
  488. data_len;
  489. __u16 mac_len,
  490. hdr_len;
  491. /* Following fields are _not_ copied in __copy_skb_header()
  492. * Note that queue_mapping is here mostly to fill a hole.
  493. */
  494. kmemcheck_bitfield_begin(flags1);
  495. __u16 queue_mapping;
  496. __u8 cloned:1,
  497. nohdr:1,
  498. fclone:2,
  499. peeked:1,
  500. head_frag:1,
  501. xmit_more:1;
  502. /* one bit hole */
  503. kmemcheck_bitfield_end(flags1);
  504. /* fields enclosed in headers_start/headers_end are copied
  505. * using a single memcpy() in __copy_skb_header()
  506. */
  507. /* private: */
  508. __u32 headers_start[0];
  509. /* public: */
  510. /* if you move pkt_type around you also must adapt those constants */
  511. #ifdef __BIG_ENDIAN_BITFIELD
  512. #define PKT_TYPE_MAX (7 << 5)
  513. #else
  514. #define PKT_TYPE_MAX 7
  515. #endif
  516. #define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
  517. __u8 __pkt_type_offset[0];
  518. __u8 pkt_type:3;
  519. __u8 pfmemalloc:1;
  520. __u8 ignore_df:1;
  521. __u8 nfctinfo:3;
  522. __u8 nf_trace:1;
  523. __u8 ip_summed:2;
  524. __u8 ooo_okay:1;
  525. __u8 l4_hash:1;
  526. __u8 sw_hash:1;
  527. __u8 wifi_acked_valid:1;
  528. __u8 wifi_acked:1;
  529. __u8 no_fcs:1;
  530. /* Indicates the inner headers are valid in the skbuff. */
  531. __u8 encapsulation:1;
  532. __u8 encap_hdr_csum:1;
  533. __u8 csum_valid:1;
  534. __u8 csum_complete_sw:1;
  535. __u8 csum_level:2;
  536. __u8 csum_bad:1;
  537. #ifdef CONFIG_IPV6_NDISC_NODETYPE
  538. __u8 ndisc_nodetype:2;
  539. #endif
  540. __u8 ipvs_property:1;
  541. __u8 inner_protocol_type:1;
  542. __u8 remcsum_offload:1;
  543. /* 3 or 5 bit hole */
  544. #ifdef CONFIG_NET_SCHED
  545. __u16 tc_index; /* traffic control index */
  546. #ifdef CONFIG_NET_CLS_ACT
  547. __u16 tc_verd; /* traffic control verdict */
  548. #endif
  549. #endif
  550. union {
  551. __wsum csum;
  552. struct {
  553. __u16 csum_start;
  554. __u16 csum_offset;
  555. };
  556. };
  557. __u32 priority;
  558. int skb_iif;
  559. __u32 hash;
  560. __be16 vlan_proto;
  561. __u16 vlan_tci;
  562. #if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
  563. union {
  564. unsigned int napi_id;
  565. unsigned int sender_cpu;
  566. };
  567. #endif
  568. #ifdef CONFIG_NETWORK_SECMARK
  569. __u32 secmark;
  570. #endif
  571. union {
  572. __u32 mark;
  573. __u32 reserved_tailroom;
  574. };
  575. union {
  576. __be16 inner_protocol;
  577. __u8 inner_ipproto;
  578. };
  579. __u16 inner_transport_header;
  580. __u16 inner_network_header;
  581. __u16 inner_mac_header;
  582. __be16 protocol;
  583. __u16 transport_header;
  584. __u16 network_header;
  585. __u16 mac_header;
  586. /* private: */
  587. __u32 headers_end[0];
  588. /* public: */
  589. /* These elements must be at the end, see alloc_skb() for details. */
  590. sk_buff_data_t tail;
  591. sk_buff_data_t end;
  592. unsigned char *head,
  593. *data;
  594. unsigned int truesize;
  595. atomic_t users;
  596. };
  597. #ifdef __KERNEL__
  598. /*
  599. * Handling routines are only of interest to the kernel
  600. */
  601. #include <linux/slab.h>
  602. #define SKB_ALLOC_FCLONE 0x01
  603. #define SKB_ALLOC_RX 0x02
  604. #define SKB_ALLOC_NAPI 0x04
  605. /* Returns true if the skb was allocated from PFMEMALLOC reserves */
  606. static inline bool skb_pfmemalloc(const struct sk_buff *skb)
  607. {
  608. return unlikely(skb->pfmemalloc);
  609. }
  610. /*
  611. * skb might have a dst pointer attached, refcounted or not.
  612. * _skb_refdst low order bit is set if refcount was _not_ taken
  613. */
  614. #define SKB_DST_NOREF 1UL
  615. #define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
  616. /**
  617. * skb_dst - returns skb dst_entry
  618. * @skb: buffer
  619. *
  620. * Returns skb dst_entry, regardless of reference taken or not.
  621. */
  622. static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
  623. {
  624. /* If refdst was not refcounted, check we still are in a
  625. * rcu_read_lock section
  626. */
  627. WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
  628. !rcu_read_lock_held() &&
  629. !rcu_read_lock_bh_held());
  630. return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
  631. }
  632. /**
  633. * skb_dst_set - sets skb dst
  634. * @skb: buffer
  635. * @dst: dst entry
  636. *
  637. * Sets skb dst, assuming a reference was taken on dst and should
  638. * be released by skb_dst_drop()
  639. */
  640. static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
  641. {
  642. skb->_skb_refdst = (unsigned long)dst;
  643. }
  644. /**
  645. * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
  646. * @skb: buffer
  647. * @dst: dst entry
  648. *
  649. * Sets skb dst, assuming a reference was not taken on dst.
  650. * If dst entry is cached, we do not take reference and dst_release
  651. * will be avoided by refdst_drop. If dst entry is not cached, we take
  652. * reference, so that last dst_release can destroy the dst immediately.
  653. */
  654. static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
  655. {
  656. WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
  657. skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
  658. }
  659. /**
  660. * skb_dst_is_noref - Test if skb dst isn't refcounted
  661. * @skb: buffer
  662. */
  663. static inline bool skb_dst_is_noref(const struct sk_buff *skb)
  664. {
  665. return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
  666. }
  667. static inline struct rtable *skb_rtable(const struct sk_buff *skb)
  668. {
  669. return (struct rtable *)skb_dst(skb);
  670. }
  671. void kfree_skb(struct sk_buff *skb);
  672. void kfree_skb_list(struct sk_buff *segs);
  673. void skb_tx_error(struct sk_buff *skb);
  674. void consume_skb(struct sk_buff *skb);
  675. void __kfree_skb(struct sk_buff *skb);
  676. extern struct kmem_cache *skbuff_head_cache;
  677. void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
  678. bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
  679. bool *fragstolen, int *delta_truesize);
  680. struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
  681. int node);
  682. struct sk_buff *__build_skb(void *data, unsigned int frag_size);
  683. struct sk_buff *build_skb(void *data, unsigned int frag_size);
  684. static inline struct sk_buff *alloc_skb(unsigned int size,
  685. gfp_t priority)
  686. {
  687. return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
  688. }
  689. struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
  690. unsigned long data_len,
  691. int max_page_order,
  692. int *errcode,
  693. gfp_t gfp_mask);
  694. /* Layout of fast clones : [skb1][skb2][fclone_ref] */
  695. struct sk_buff_fclones {
  696. struct sk_buff skb1;
  697. struct sk_buff skb2;
  698. atomic_t fclone_ref;
  699. };
  700. /**
  701. * skb_fclone_busy - check if fclone is busy
  702. * @skb: buffer
  703. *
  704. * Returns true is skb is a fast clone, and its clone is not freed.
  705. * Some drivers call skb_orphan() in their ndo_start_xmit(),
  706. * so we also check that this didnt happen.
  707. */
  708. static inline bool skb_fclone_busy(const struct sock *sk,
  709. const struct sk_buff *skb)
  710. {
  711. const struct sk_buff_fclones *fclones;
  712. fclones = container_of(skb, struct sk_buff_fclones, skb1);
  713. return skb->fclone == SKB_FCLONE_ORIG &&
  714. atomic_read(&fclones->fclone_ref) > 1 &&
  715. fclones->skb2.sk == sk;
  716. }
  717. static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
  718. gfp_t priority)
  719. {
  720. return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
  721. }
  722. struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
  723. static inline struct sk_buff *alloc_skb_head(gfp_t priority)
  724. {
  725. return __alloc_skb_head(priority, -1);
  726. }
  727. struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
  728. int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
  729. struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
  730. struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
  731. struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
  732. gfp_t gfp_mask, bool fclone);
  733. static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
  734. gfp_t gfp_mask)
  735. {
  736. return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
  737. }
  738. int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
  739. struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
  740. unsigned int headroom);
  741. struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
  742. int newtailroom, gfp_t priority);
  743. int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
  744. int offset, int len);
  745. int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
  746. int len);
  747. int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
  748. int skb_pad(struct sk_buff *skb, int pad);
  749. #define dev_kfree_skb(a) consume_skb(a)
  750. int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
  751. int getfrag(void *from, char *to, int offset,
  752. int len, int odd, struct sk_buff *skb),
  753. void *from, int length);
  754. struct skb_seq_state {
  755. __u32 lower_offset;
  756. __u32 upper_offset;
  757. __u32 frag_idx;
  758. __u32 stepped_offset;
  759. struct sk_buff *root_skb;
  760. struct sk_buff *cur_skb;
  761. __u8 *frag_data;
  762. };
  763. void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
  764. unsigned int to, struct skb_seq_state *st);
  765. unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
  766. struct skb_seq_state *st);
  767. void skb_abort_seq_read(struct skb_seq_state *st);
  768. unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
  769. unsigned int to, struct ts_config *config);
  770. /*
  771. * Packet hash types specify the type of hash in skb_set_hash.
  772. *
  773. * Hash types refer to the protocol layer addresses which are used to
  774. * construct a packet's hash. The hashes are used to differentiate or identify
  775. * flows of the protocol layer for the hash type. Hash types are either
  776. * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
  777. *
  778. * Properties of hashes:
  779. *
  780. * 1) Two packets in different flows have different hash values
  781. * 2) Two packets in the same flow should have the same hash value
  782. *
  783. * A hash at a higher layer is considered to be more specific. A driver should
  784. * set the most specific hash possible.
  785. *
  786. * A driver cannot indicate a more specific hash than the layer at which a hash
  787. * was computed. For instance an L3 hash cannot be set as an L4 hash.
  788. *
  789. * A driver may indicate a hash level which is less specific than the
  790. * actual layer the hash was computed on. For instance, a hash computed
  791. * at L4 may be considered an L3 hash. This should only be done if the
  792. * driver can't unambiguously determine that the HW computed the hash at
  793. * the higher layer. Note that the "should" in the second property above
  794. * permits this.
  795. */
  796. enum pkt_hash_types {
  797. PKT_HASH_TYPE_NONE, /* Undefined type */
  798. PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
  799. PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
  800. PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
  801. };
  802. static inline void
  803. skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
  804. {
  805. skb->l4_hash = (type == PKT_HASH_TYPE_L4);
  806. skb->sw_hash = 0;
  807. skb->hash = hash;
  808. }
  809. void __skb_get_hash(struct sk_buff *skb);
  810. static inline __u32 skb_get_hash(struct sk_buff *skb)
  811. {
  812. if (!skb->l4_hash && !skb->sw_hash)
  813. __skb_get_hash(skb);
  814. return skb->hash;
  815. }
  816. static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
  817. {
  818. return skb->hash;
  819. }
  820. static inline void skb_clear_hash(struct sk_buff *skb)
  821. {
  822. skb->hash = 0;
  823. skb->sw_hash = 0;
  824. skb->l4_hash = 0;
  825. }
  826. static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
  827. {
  828. if (!skb->l4_hash)
  829. skb_clear_hash(skb);
  830. }
  831. static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
  832. {
  833. to->hash = from->hash;
  834. to->sw_hash = from->sw_hash;
  835. to->l4_hash = from->l4_hash;
  836. };
  837. static inline void skb_sender_cpu_clear(struct sk_buff *skb)
  838. {
  839. #ifdef CONFIG_XPS
  840. skb->sender_cpu = 0;
  841. #endif
  842. }
  843. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  844. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  845. {
  846. return skb->head + skb->end;
  847. }
  848. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  849. {
  850. return skb->end;
  851. }
  852. #else
  853. static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
  854. {
  855. return skb->end;
  856. }
  857. static inline unsigned int skb_end_offset(const struct sk_buff *skb)
  858. {
  859. return skb->end - skb->head;
  860. }
  861. #endif
  862. /* Internal */
  863. #define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
  864. static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
  865. {
  866. return &skb_shinfo(skb)->hwtstamps;
  867. }
  868. /**
  869. * skb_queue_empty - check if a queue is empty
  870. * @list: queue head
  871. *
  872. * Returns true if the queue is empty, false otherwise.
  873. */
  874. static inline int skb_queue_empty(const struct sk_buff_head *list)
  875. {
  876. return list->next == (const struct sk_buff *) list;
  877. }
  878. /**
  879. * skb_queue_is_last - check if skb is the last entry in the queue
  880. * @list: queue head
  881. * @skb: buffer
  882. *
  883. * Returns true if @skb is the last buffer on the list.
  884. */
  885. static inline bool skb_queue_is_last(const struct sk_buff_head *list,
  886. const struct sk_buff *skb)
  887. {
  888. return skb->next == (const struct sk_buff *) list;
  889. }
  890. /**
  891. * skb_queue_is_first - check if skb is the first entry in the queue
  892. * @list: queue head
  893. * @skb: buffer
  894. *
  895. * Returns true if @skb is the first buffer on the list.
  896. */
  897. static inline bool skb_queue_is_first(const struct sk_buff_head *list,
  898. const struct sk_buff *skb)
  899. {
  900. return skb->prev == (const struct sk_buff *) list;
  901. }
  902. /**
  903. * skb_queue_next - return the next packet in the queue
  904. * @list: queue head
  905. * @skb: current buffer
  906. *
  907. * Return the next packet in @list after @skb. It is only valid to
  908. * call this if skb_queue_is_last() evaluates to false.
  909. */
  910. static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
  911. const struct sk_buff *skb)
  912. {
  913. /* This BUG_ON may seem severe, but if we just return then we
  914. * are going to dereference garbage.
  915. */
  916. BUG_ON(skb_queue_is_last(list, skb));
  917. return skb->next;
  918. }
  919. /**
  920. * skb_queue_prev - return the prev packet in the queue
  921. * @list: queue head
  922. * @skb: current buffer
  923. *
  924. * Return the prev packet in @list before @skb. It is only valid to
  925. * call this if skb_queue_is_first() evaluates to false.
  926. */
  927. static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
  928. const struct sk_buff *skb)
  929. {
  930. /* This BUG_ON may seem severe, but if we just return then we
  931. * are going to dereference garbage.
  932. */
  933. BUG_ON(skb_queue_is_first(list, skb));
  934. return skb->prev;
  935. }
  936. /**
  937. * skb_get - reference buffer
  938. * @skb: buffer to reference
  939. *
  940. * Makes another reference to a socket buffer and returns a pointer
  941. * to the buffer.
  942. */
  943. static inline struct sk_buff *skb_get(struct sk_buff *skb)
  944. {
  945. atomic_inc(&skb->users);
  946. return skb;
  947. }
  948. /*
  949. * If users == 1, we are the only owner and are can avoid redundant
  950. * atomic change.
  951. */
  952. /**
  953. * skb_cloned - is the buffer a clone
  954. * @skb: buffer to check
  955. *
  956. * Returns true if the buffer was generated with skb_clone() and is
  957. * one of multiple shared copies of the buffer. Cloned buffers are
  958. * shared data so must not be written to under normal circumstances.
  959. */
  960. static inline int skb_cloned(const struct sk_buff *skb)
  961. {
  962. return skb->cloned &&
  963. (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
  964. }
  965. static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
  966. {
  967. might_sleep_if(pri & __GFP_WAIT);
  968. if (skb_cloned(skb))
  969. return pskb_expand_head(skb, 0, 0, pri);
  970. return 0;
  971. }
  972. /**
  973. * skb_header_cloned - is the header a clone
  974. * @skb: buffer to check
  975. *
  976. * Returns true if modifying the header part of the buffer requires
  977. * the data to be copied.
  978. */
  979. static inline int skb_header_cloned(const struct sk_buff *skb)
  980. {
  981. int dataref;
  982. if (!skb->cloned)
  983. return 0;
  984. dataref = atomic_read(&skb_shinfo(skb)->dataref);
  985. dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
  986. return dataref != 1;
  987. }
  988. /**
  989. * skb_header_release - release reference to header
  990. * @skb: buffer to operate on
  991. *
  992. * Drop a reference to the header part of the buffer. This is done
  993. * by acquiring a payload reference. You must not read from the header
  994. * part of skb->data after this.
  995. * Note : Check if you can use __skb_header_release() instead.
  996. */
  997. static inline void skb_header_release(struct sk_buff *skb)
  998. {
  999. BUG_ON(skb->nohdr);
  1000. skb->nohdr = 1;
  1001. atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
  1002. }
  1003. /**
  1004. * __skb_header_release - release reference to header
  1005. * @skb: buffer to operate on
  1006. *
  1007. * Variant of skb_header_release() assuming skb is private to caller.
  1008. * We can avoid one atomic operation.
  1009. */
  1010. static inline void __skb_header_release(struct sk_buff *skb)
  1011. {
  1012. skb->nohdr = 1;
  1013. atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
  1014. }
  1015. /**
  1016. * skb_shared - is the buffer shared
  1017. * @skb: buffer to check
  1018. *
  1019. * Returns true if more than one person has a reference to this
  1020. * buffer.
  1021. */
  1022. static inline int skb_shared(const struct sk_buff *skb)
  1023. {
  1024. return atomic_read(&skb->users) != 1;
  1025. }
  1026. /**
  1027. * skb_share_check - check if buffer is shared and if so clone it
  1028. * @skb: buffer to check
  1029. * @pri: priority for memory allocation
  1030. *
  1031. * If the buffer is shared the buffer is cloned and the old copy
  1032. * drops a reference. A new clone with a single reference is returned.
  1033. * If the buffer is not shared the original buffer is returned. When
  1034. * being called from interrupt status or with spinlocks held pri must
  1035. * be GFP_ATOMIC.
  1036. *
  1037. * NULL is returned on a memory allocation failure.
  1038. */
  1039. static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
  1040. {
  1041. might_sleep_if(pri & __GFP_WAIT);
  1042. if (skb_shared(skb)) {
  1043. struct sk_buff *nskb = skb_clone(skb, pri);
  1044. if (likely(nskb))
  1045. consume_skb(skb);
  1046. else
  1047. kfree_skb(skb);
  1048. skb = nskb;
  1049. }
  1050. return skb;
  1051. }
  1052. /*
  1053. * Copy shared buffers into a new sk_buff. We effectively do COW on
  1054. * packets to handle cases where we have a local reader and forward
  1055. * and a couple of other messy ones. The normal one is tcpdumping
  1056. * a packet thats being forwarded.
  1057. */
  1058. /**
  1059. * skb_unshare - make a copy of a shared buffer
  1060. * @skb: buffer to check
  1061. * @pri: priority for memory allocation
  1062. *
  1063. * If the socket buffer is a clone then this function creates a new
  1064. * copy of the data, drops a reference count on the old copy and returns
  1065. * the new copy with the reference count at 1. If the buffer is not a clone
  1066. * the original buffer is returned. When called with a spinlock held or
  1067. * from interrupt state @pri must be %GFP_ATOMIC
  1068. *
  1069. * %NULL is returned on a memory allocation failure.
  1070. */
  1071. static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
  1072. gfp_t pri)
  1073. {
  1074. might_sleep_if(pri & __GFP_WAIT);
  1075. if (skb_cloned(skb)) {
  1076. struct sk_buff *nskb = skb_copy(skb, pri);
  1077. /* Free our shared copy */
  1078. if (likely(nskb))
  1079. consume_skb(skb);
  1080. else
  1081. kfree_skb(skb);
  1082. skb = nskb;
  1083. }
  1084. return skb;
  1085. }
  1086. /**
  1087. * skb_peek - peek at the head of an &sk_buff_head
  1088. * @list_: list to peek at
  1089. *
  1090. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1091. * be careful with this one. A peek leaves the buffer on the
  1092. * list and someone else may run off with it. You must hold
  1093. * the appropriate locks or have a private queue to do this.
  1094. *
  1095. * Returns %NULL for an empty list or a pointer to the head element.
  1096. * The reference count is not incremented and the reference is therefore
  1097. * volatile. Use with caution.
  1098. */
  1099. static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
  1100. {
  1101. struct sk_buff *skb = list_->next;
  1102. if (skb == (struct sk_buff *)list_)
  1103. skb = NULL;
  1104. return skb;
  1105. }
  1106. /**
  1107. * skb_peek_next - peek skb following the given one from a queue
  1108. * @skb: skb to start from
  1109. * @list_: list to peek at
  1110. *
  1111. * Returns %NULL when the end of the list is met or a pointer to the
  1112. * next element. The reference count is not incremented and the
  1113. * reference is therefore volatile. Use with caution.
  1114. */
  1115. static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
  1116. const struct sk_buff_head *list_)
  1117. {
  1118. struct sk_buff *next = skb->next;
  1119. if (next == (struct sk_buff *)list_)
  1120. next = NULL;
  1121. return next;
  1122. }
  1123. /**
  1124. * skb_peek_tail - peek at the tail of an &sk_buff_head
  1125. * @list_: list to peek at
  1126. *
  1127. * Peek an &sk_buff. Unlike most other operations you _MUST_
  1128. * be careful with this one. A peek leaves the buffer on the
  1129. * list and someone else may run off with it. You must hold
  1130. * the appropriate locks or have a private queue to do this.
  1131. *
  1132. * Returns %NULL for an empty list or a pointer to the tail element.
  1133. * The reference count is not incremented and the reference is therefore
  1134. * volatile. Use with caution.
  1135. */
  1136. static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
  1137. {
  1138. struct sk_buff *skb = list_->prev;
  1139. if (skb == (struct sk_buff *)list_)
  1140. skb = NULL;
  1141. return skb;
  1142. }
  1143. /**
  1144. * skb_queue_len - get queue length
  1145. * @list_: list to measure
  1146. *
  1147. * Return the length of an &sk_buff queue.
  1148. */
  1149. static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
  1150. {
  1151. return list_->qlen;
  1152. }
  1153. /**
  1154. * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
  1155. * @list: queue to initialize
  1156. *
  1157. * This initializes only the list and queue length aspects of
  1158. * an sk_buff_head object. This allows to initialize the list
  1159. * aspects of an sk_buff_head without reinitializing things like
  1160. * the spinlock. It can also be used for on-stack sk_buff_head
  1161. * objects where the spinlock is known to not be used.
  1162. */
  1163. static inline void __skb_queue_head_init(struct sk_buff_head *list)
  1164. {
  1165. list->prev = list->next = (struct sk_buff *)list;
  1166. list->qlen = 0;
  1167. }
  1168. /*
  1169. * This function creates a split out lock class for each invocation;
  1170. * this is needed for now since a whole lot of users of the skb-queue
  1171. * infrastructure in drivers have different locking usage (in hardirq)
  1172. * than the networking core (in softirq only). In the long run either the
  1173. * network layer or drivers should need annotation to consolidate the
  1174. * main types of usage into 3 classes.
  1175. */
  1176. static inline void skb_queue_head_init(struct sk_buff_head *list)
  1177. {
  1178. spin_lock_init(&list->lock);
  1179. __skb_queue_head_init(list);
  1180. }
  1181. static inline void skb_queue_head_init_class(struct sk_buff_head *list,
  1182. struct lock_class_key *class)
  1183. {
  1184. skb_queue_head_init(list);
  1185. lockdep_set_class(&list->lock, class);
  1186. }
  1187. /*
  1188. * Insert an sk_buff on a list.
  1189. *
  1190. * The "__skb_xxxx()" functions are the non-atomic ones that
  1191. * can only be called with interrupts disabled.
  1192. */
  1193. void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
  1194. struct sk_buff_head *list);
  1195. static inline void __skb_insert(struct sk_buff *newsk,
  1196. struct sk_buff *prev, struct sk_buff *next,
  1197. struct sk_buff_head *list)
  1198. {
  1199. newsk->next = next;
  1200. newsk->prev = prev;
  1201. next->prev = prev->next = newsk;
  1202. list->qlen++;
  1203. }
  1204. static inline void __skb_queue_splice(const struct sk_buff_head *list,
  1205. struct sk_buff *prev,
  1206. struct sk_buff *next)
  1207. {
  1208. struct sk_buff *first = list->next;
  1209. struct sk_buff *last = list->prev;
  1210. first->prev = prev;
  1211. prev->next = first;
  1212. last->next = next;
  1213. next->prev = last;
  1214. }
  1215. /**
  1216. * skb_queue_splice - join two skb lists, this is designed for stacks
  1217. * @list: the new list to add
  1218. * @head: the place to add it in the first list
  1219. */
  1220. static inline void skb_queue_splice(const struct sk_buff_head *list,
  1221. struct sk_buff_head *head)
  1222. {
  1223. if (!skb_queue_empty(list)) {
  1224. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1225. head->qlen += list->qlen;
  1226. }
  1227. }
  1228. /**
  1229. * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
  1230. * @list: the new list to add
  1231. * @head: the place to add it in the first list
  1232. *
  1233. * The list at @list is reinitialised
  1234. */
  1235. static inline void skb_queue_splice_init(struct sk_buff_head *list,
  1236. struct sk_buff_head *head)
  1237. {
  1238. if (!skb_queue_empty(list)) {
  1239. __skb_queue_splice(list, (struct sk_buff *) head, head->next);
  1240. head->qlen += list->qlen;
  1241. __skb_queue_head_init(list);
  1242. }
  1243. }
  1244. /**
  1245. * skb_queue_splice_tail - join two skb lists, each list being a queue
  1246. * @list: the new list to add
  1247. * @head: the place to add it in the first list
  1248. */
  1249. static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
  1250. struct sk_buff_head *head)
  1251. {
  1252. if (!skb_queue_empty(list)) {
  1253. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1254. head->qlen += list->qlen;
  1255. }
  1256. }
  1257. /**
  1258. * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
  1259. * @list: the new list to add
  1260. * @head: the place to add it in the first list
  1261. *
  1262. * Each of the lists is a queue.
  1263. * The list at @list is reinitialised
  1264. */
  1265. static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
  1266. struct sk_buff_head *head)
  1267. {
  1268. if (!skb_queue_empty(list)) {
  1269. __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
  1270. head->qlen += list->qlen;
  1271. __skb_queue_head_init(list);
  1272. }
  1273. }
  1274. /**
  1275. * __skb_queue_after - queue a buffer at the list head
  1276. * @list: list to use
  1277. * @prev: place after this buffer
  1278. * @newsk: buffer to queue
  1279. *
  1280. * Queue a buffer int the middle of a list. This function takes no locks
  1281. * and you must therefore hold required locks before calling it.
  1282. *
  1283. * A buffer cannot be placed on two lists at the same time.
  1284. */
  1285. static inline void __skb_queue_after(struct sk_buff_head *list,
  1286. struct sk_buff *prev,
  1287. struct sk_buff *newsk)
  1288. {
  1289. __skb_insert(newsk, prev, prev->next, list);
  1290. }
  1291. void skb_append(struct sk_buff *old, struct sk_buff *newsk,
  1292. struct sk_buff_head *list);
  1293. static inline void __skb_queue_before(struct sk_buff_head *list,
  1294. struct sk_buff *next,
  1295. struct sk_buff *newsk)
  1296. {
  1297. __skb_insert(newsk, next->prev, next, list);
  1298. }
  1299. /**
  1300. * __skb_queue_head - queue a buffer at the list head
  1301. * @list: list to use
  1302. * @newsk: buffer to queue
  1303. *
  1304. * Queue a buffer at the start of a list. This function takes no locks
  1305. * and you must therefore hold required locks before calling it.
  1306. *
  1307. * A buffer cannot be placed on two lists at the same time.
  1308. */
  1309. void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
  1310. static inline void __skb_queue_head(struct sk_buff_head *list,
  1311. struct sk_buff *newsk)
  1312. {
  1313. __skb_queue_after(list, (struct sk_buff *)list, newsk);
  1314. }
  1315. /**
  1316. * __skb_queue_tail - queue a buffer at the list tail
  1317. * @list: list to use
  1318. * @newsk: buffer to queue
  1319. *
  1320. * Queue a buffer at the end of a list. This function takes no locks
  1321. * and you must therefore hold required locks before calling it.
  1322. *
  1323. * A buffer cannot be placed on two lists at the same time.
  1324. */
  1325. void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
  1326. static inline void __skb_queue_tail(struct sk_buff_head *list,
  1327. struct sk_buff *newsk)
  1328. {
  1329. __skb_queue_before(list, (struct sk_buff *)list, newsk);
  1330. }
  1331. /*
  1332. * remove sk_buff from list. _Must_ be called atomically, and with
  1333. * the list known..
  1334. */
  1335. void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
  1336. static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
  1337. {
  1338. struct sk_buff *next, *prev;
  1339. list->qlen--;
  1340. next = skb->next;
  1341. prev = skb->prev;
  1342. skb->next = skb->prev = NULL;
  1343. next->prev = prev;
  1344. prev->next = next;
  1345. }
  1346. /**
  1347. * __skb_dequeue - remove from the head of the queue
  1348. * @list: list to dequeue from
  1349. *
  1350. * Remove the head of the list. This function does not take any locks
  1351. * so must be used with appropriate locks held only. The head item is
  1352. * returned or %NULL if the list is empty.
  1353. */
  1354. struct sk_buff *skb_dequeue(struct sk_buff_head *list);
  1355. static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
  1356. {
  1357. struct sk_buff *skb = skb_peek(list);
  1358. if (skb)
  1359. __skb_unlink(skb, list);
  1360. return skb;
  1361. }
  1362. /**
  1363. * __skb_dequeue_tail - remove from the tail of the queue
  1364. * @list: list to dequeue from
  1365. *
  1366. * Remove the tail of the list. This function does not take any locks
  1367. * so must be used with appropriate locks held only. The tail item is
  1368. * returned or %NULL if the list is empty.
  1369. */
  1370. struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
  1371. static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
  1372. {
  1373. struct sk_buff *skb = skb_peek_tail(list);
  1374. if (skb)
  1375. __skb_unlink(skb, list);
  1376. return skb;
  1377. }
  1378. static inline bool skb_is_nonlinear(const struct sk_buff *skb)
  1379. {
  1380. return skb->data_len;
  1381. }
  1382. static inline unsigned int skb_headlen(const struct sk_buff *skb)
  1383. {
  1384. return skb->len - skb->data_len;
  1385. }
  1386. static inline int skb_pagelen(const struct sk_buff *skb)
  1387. {
  1388. int i, len = 0;
  1389. for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
  1390. len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
  1391. return len + skb_headlen(skb);
  1392. }
  1393. /**
  1394. * __skb_fill_page_desc - initialise a paged fragment in an skb
  1395. * @skb: buffer containing fragment to be initialised
  1396. * @i: paged fragment index to initialise
  1397. * @page: the page to use for this fragment
  1398. * @off: the offset to the data with @page
  1399. * @size: the length of the data
  1400. *
  1401. * Initialises the @i'th fragment of @skb to point to &size bytes at
  1402. * offset @off within @page.
  1403. *
  1404. * Does not take any additional reference on the fragment.
  1405. */
  1406. static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
  1407. struct page *page, int off, int size)
  1408. {
  1409. skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
  1410. /*
  1411. * Propagate page->pfmemalloc to the skb if we can. The problem is
  1412. * that not all callers have unique ownership of the page. If
  1413. * pfmemalloc is set, we check the mapping as a mapping implies
  1414. * page->index is set (index and pfmemalloc share space).
  1415. * If it's a valid mapping, we cannot use page->pfmemalloc but we
  1416. * do not lose pfmemalloc information as the pages would not be
  1417. * allocated using __GFP_MEMALLOC.
  1418. */
  1419. frag->page.p = page;
  1420. frag->page_offset = off;
  1421. skb_frag_size_set(frag, size);
  1422. page = compound_head(page);
  1423. if (page->pfmemalloc && !page->mapping)
  1424. skb->pfmemalloc = true;
  1425. }
  1426. /**
  1427. * skb_fill_page_desc - initialise a paged fragment in an skb
  1428. * @skb: buffer containing fragment to be initialised
  1429. * @i: paged fragment index to initialise
  1430. * @page: the page to use for this fragment
  1431. * @off: the offset to the data with @page
  1432. * @size: the length of the data
  1433. *
  1434. * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
  1435. * @skb to point to @size bytes at offset @off within @page. In
  1436. * addition updates @skb such that @i is the last fragment.
  1437. *
  1438. * Does not take any additional reference on the fragment.
  1439. */
  1440. static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
  1441. struct page *page, int off, int size)
  1442. {
  1443. __skb_fill_page_desc(skb, i, page, off, size);
  1444. skb_shinfo(skb)->nr_frags = i + 1;
  1445. }
  1446. void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
  1447. int size, unsigned int truesize);
  1448. void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
  1449. unsigned int truesize);
  1450. #define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
  1451. #define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
  1452. #define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
  1453. #ifdef NET_SKBUFF_DATA_USES_OFFSET
  1454. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1455. {
  1456. return skb->head + skb->tail;
  1457. }
  1458. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1459. {
  1460. skb->tail = skb->data - skb->head;
  1461. }
  1462. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1463. {
  1464. skb_reset_tail_pointer(skb);
  1465. skb->tail += offset;
  1466. }
  1467. #else /* NET_SKBUFF_DATA_USES_OFFSET */
  1468. static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
  1469. {
  1470. return skb->tail;
  1471. }
  1472. static inline void skb_reset_tail_pointer(struct sk_buff *skb)
  1473. {
  1474. skb->tail = skb->data;
  1475. }
  1476. static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
  1477. {
  1478. skb->tail = skb->data + offset;
  1479. }
  1480. #endif /* NET_SKBUFF_DATA_USES_OFFSET */
  1481. /*
  1482. * Add data to an sk_buff
  1483. */
  1484. unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
  1485. unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
  1486. static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
  1487. {
  1488. unsigned char *tmp = skb_tail_pointer(skb);
  1489. SKB_LINEAR_ASSERT(skb);
  1490. skb->tail += len;
  1491. skb->len += len;
  1492. return tmp;
  1493. }
  1494. unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
  1495. static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
  1496. {
  1497. skb->data -= len;
  1498. skb->len += len;
  1499. return skb->data;
  1500. }
  1501. unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
  1502. static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
  1503. {
  1504. skb->len -= len;
  1505. BUG_ON(skb->len < skb->data_len);
  1506. return skb->data += len;
  1507. }
  1508. static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
  1509. {
  1510. return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
  1511. }
  1512. unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
  1513. static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
  1514. {
  1515. if (len > skb_headlen(skb) &&
  1516. !__pskb_pull_tail(skb, len - skb_headlen(skb)))
  1517. return NULL;
  1518. skb->len -= len;
  1519. return skb->data += len;
  1520. }
  1521. static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
  1522. {
  1523. return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
  1524. }
  1525. static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
  1526. {
  1527. if (likely(len <= skb_headlen(skb)))
  1528. return 1;
  1529. if (unlikely(len > skb->len))
  1530. return 0;
  1531. return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
  1532. }
  1533. /**
  1534. * skb_headroom - bytes at buffer head
  1535. * @skb: buffer to check
  1536. *
  1537. * Return the number of bytes of free space at the head of an &sk_buff.
  1538. */
  1539. static inline unsigned int skb_headroom(const struct sk_buff *skb)
  1540. {
  1541. return skb->data - skb->head;
  1542. }
  1543. /**
  1544. * skb_tailroom - bytes at buffer end
  1545. * @skb: buffer to check
  1546. *
  1547. * Return the number of bytes of free space at the tail of an sk_buff
  1548. */
  1549. static inline int skb_tailroom(const struct sk_buff *skb)
  1550. {
  1551. return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
  1552. }
  1553. /**
  1554. * skb_availroom - bytes at buffer end
  1555. * @skb: buffer to check
  1556. *
  1557. * Return the number of bytes of free space at the tail of an sk_buff
  1558. * allocated by sk_stream_alloc()
  1559. */
  1560. static inline int skb_availroom(const struct sk_buff *skb)
  1561. {
  1562. if (skb_is_nonlinear(skb))
  1563. return 0;
  1564. return skb->end - skb->tail - skb->reserved_tailroom;
  1565. }
  1566. /**
  1567. * skb_reserve - adjust headroom
  1568. * @skb: buffer to alter
  1569. * @len: bytes to move
  1570. *
  1571. * Increase the headroom of an empty &sk_buff by reducing the tail
  1572. * room. This is only allowed for an empty buffer.
  1573. */
  1574. static inline void skb_reserve(struct sk_buff *skb, int len)
  1575. {
  1576. skb->data += len;
  1577. skb->tail += len;
  1578. }
  1579. #define ENCAP_TYPE_ETHER 0
  1580. #define ENCAP_TYPE_IPPROTO 1
  1581. static inline void skb_set_inner_protocol(struct sk_buff *skb,
  1582. __be16 protocol)
  1583. {
  1584. skb->inner_protocol = protocol;
  1585. skb->inner_protocol_type = ENCAP_TYPE_ETHER;
  1586. }
  1587. static inline void skb_set_inner_ipproto(struct sk_buff *skb,
  1588. __u8 ipproto)
  1589. {
  1590. skb->inner_ipproto = ipproto;
  1591. skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
  1592. }
  1593. static inline void skb_reset_inner_headers(struct sk_buff *skb)
  1594. {
  1595. skb->inner_mac_header = skb->mac_header;
  1596. skb->inner_network_header = skb->network_header;
  1597. skb->inner_transport_header = skb->transport_header;
  1598. }
  1599. static inline void skb_reset_mac_len(struct sk_buff *skb)
  1600. {
  1601. skb->mac_len = skb->network_header - skb->mac_header;
  1602. }
  1603. static inline unsigned char *skb_inner_transport_header(const struct sk_buff
  1604. *skb)
  1605. {
  1606. return skb->head + skb->inner_transport_header;
  1607. }
  1608. static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
  1609. {
  1610. skb->inner_transport_header = skb->data - skb->head;
  1611. }
  1612. static inline void skb_set_inner_transport_header(struct sk_buff *skb,
  1613. const int offset)
  1614. {
  1615. skb_reset_inner_transport_header(skb);
  1616. skb->inner_transport_header += offset;
  1617. }
  1618. static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
  1619. {
  1620. return skb->head + skb->inner_network_header;
  1621. }
  1622. static inline void skb_reset_inner_network_header(struct sk_buff *skb)
  1623. {
  1624. skb->inner_network_header = skb->data - skb->head;
  1625. }
  1626. static inline void skb_set_inner_network_header(struct sk_buff *skb,
  1627. const int offset)
  1628. {
  1629. skb_reset_inner_network_header(skb);
  1630. skb->inner_network_header += offset;
  1631. }
  1632. static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
  1633. {
  1634. return skb->head + skb->inner_mac_header;
  1635. }
  1636. static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
  1637. {
  1638. skb->inner_mac_header = skb->data - skb->head;
  1639. }
  1640. static inline void skb_set_inner_mac_header(struct sk_buff *skb,
  1641. const int offset)
  1642. {
  1643. skb_reset_inner_mac_header(skb);
  1644. skb->inner_mac_header += offset;
  1645. }
  1646. static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
  1647. {
  1648. return skb->transport_header != (typeof(skb->transport_header))~0U;
  1649. }
  1650. static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
  1651. {
  1652. return skb->head + skb->transport_header;
  1653. }
  1654. static inline void skb_reset_transport_header(struct sk_buff *skb)
  1655. {
  1656. skb->transport_header = skb->data - skb->head;
  1657. }
  1658. static inline void skb_set_transport_header(struct sk_buff *skb,
  1659. const int offset)
  1660. {
  1661. skb_reset_transport_header(skb);
  1662. skb->transport_header += offset;
  1663. }
  1664. static inline unsigned char *skb_network_header(const struct sk_buff *skb)
  1665. {
  1666. return skb->head + skb->network_header;
  1667. }
  1668. static inline void skb_reset_network_header(struct sk_buff *skb)
  1669. {
  1670. skb->network_header = skb->data - skb->head;
  1671. }
  1672. static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
  1673. {
  1674. skb_reset_network_header(skb);
  1675. skb->network_header += offset;
  1676. }
  1677. static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
  1678. {
  1679. return skb->head + skb->mac_header;
  1680. }
  1681. static inline int skb_mac_header_was_set(const struct sk_buff *skb)
  1682. {
  1683. return skb->mac_header != (typeof(skb->mac_header))~0U;
  1684. }
  1685. static inline void skb_reset_mac_header(struct sk_buff *skb)
  1686. {
  1687. skb->mac_header = skb->data - skb->head;
  1688. }
  1689. static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
  1690. {
  1691. skb_reset_mac_header(skb);
  1692. skb->mac_header += offset;
  1693. }
  1694. static inline void skb_pop_mac_header(struct sk_buff *skb)
  1695. {
  1696. skb->mac_header = skb->network_header;
  1697. }
  1698. static inline void skb_probe_transport_header(struct sk_buff *skb,
  1699. const int offset_hint)
  1700. {
  1701. struct flow_keys keys;
  1702. if (skb_transport_header_was_set(skb))
  1703. return;
  1704. else if (skb_flow_dissect(skb, &keys))
  1705. skb_set_transport_header(skb, keys.thoff);
  1706. else
  1707. skb_set_transport_header(skb, offset_hint);
  1708. }
  1709. static inline void skb_mac_header_rebuild(struct sk_buff *skb)
  1710. {
  1711. if (skb_mac_header_was_set(skb)) {
  1712. const unsigned char *old_mac = skb_mac_header(skb);
  1713. skb_set_mac_header(skb, -skb->mac_len);
  1714. memmove(skb_mac_header(skb), old_mac, skb->mac_len);
  1715. }
  1716. }
  1717. static inline int skb_checksum_start_offset(const struct sk_buff *skb)
  1718. {
  1719. return skb->csum_start - skb_headroom(skb);
  1720. }
  1721. static inline int skb_transport_offset(const struct sk_buff *skb)
  1722. {
  1723. return skb_transport_header(skb) - skb->data;
  1724. }
  1725. static inline u32 skb_network_header_len(const struct sk_buff *skb)
  1726. {
  1727. return skb->transport_header - skb->network_header;
  1728. }
  1729. static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
  1730. {
  1731. return skb->inner_transport_header - skb->inner_network_header;
  1732. }
  1733. static inline int skb_network_offset(const struct sk_buff *skb)
  1734. {
  1735. return skb_network_header(skb) - skb->data;
  1736. }
  1737. static inline int skb_inner_network_offset(const struct sk_buff *skb)
  1738. {
  1739. return skb_inner_network_header(skb) - skb->data;
  1740. }
  1741. static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
  1742. {
  1743. return pskb_may_pull(skb, skb_network_offset(skb) + len);
  1744. }
  1745. /*
  1746. * CPUs often take a performance hit when accessing unaligned memory
  1747. * locations. The actual performance hit varies, it can be small if the
  1748. * hardware handles it or large if we have to take an exception and fix it
  1749. * in software.
  1750. *
  1751. * Since an ethernet header is 14 bytes network drivers often end up with
  1752. * the IP header at an unaligned offset. The IP header can be aligned by
  1753. * shifting the start of the packet by 2 bytes. Drivers should do this
  1754. * with:
  1755. *
  1756. * skb_reserve(skb, NET_IP_ALIGN);
  1757. *
  1758. * The downside to this alignment of the IP header is that the DMA is now
  1759. * unaligned. On some architectures the cost of an unaligned DMA is high
  1760. * and this cost outweighs the gains made by aligning the IP header.
  1761. *
  1762. * Since this trade off varies between architectures, we allow NET_IP_ALIGN
  1763. * to be overridden.
  1764. */
  1765. #ifndef NET_IP_ALIGN
  1766. #define NET_IP_ALIGN 2
  1767. #endif
  1768. /*
  1769. * The networking layer reserves some headroom in skb data (via
  1770. * dev_alloc_skb). This is used to avoid having to reallocate skb data when
  1771. * the header has to grow. In the default case, if the header has to grow
  1772. * 32 bytes or less we avoid the reallocation.
  1773. *
  1774. * Unfortunately this headroom changes the DMA alignment of the resulting
  1775. * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
  1776. * on some architectures. An architecture can override this value,
  1777. * perhaps setting it to a cacheline in size (since that will maintain
  1778. * cacheline alignment of the DMA). It must be a power of 2.
  1779. *
  1780. * Various parts of the networking layer expect at least 32 bytes of
  1781. * headroom, you should not reduce this.
  1782. *
  1783. * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
  1784. * to reduce average number of cache lines per packet.
  1785. * get_rps_cpus() for example only access one 64 bytes aligned block :
  1786. * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
  1787. */
  1788. #ifndef NET_SKB_PAD
  1789. #define NET_SKB_PAD max(32, L1_CACHE_BYTES)
  1790. #endif
  1791. int ___pskb_trim(struct sk_buff *skb, unsigned int len);
  1792. static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
  1793. {
  1794. if (unlikely(skb_is_nonlinear(skb))) {
  1795. WARN_ON(1);
  1796. return;
  1797. }
  1798. skb->len = len;
  1799. skb_set_tail_pointer(skb, len);
  1800. }
  1801. void skb_trim(struct sk_buff *skb, unsigned int len);
  1802. static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
  1803. {
  1804. if (skb->data_len)
  1805. return ___pskb_trim(skb, len);
  1806. __skb_trim(skb, len);
  1807. return 0;
  1808. }
  1809. static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
  1810. {
  1811. return (len < skb->len) ? __pskb_trim(skb, len) : 0;
  1812. }
  1813. /**
  1814. * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
  1815. * @skb: buffer to alter
  1816. * @len: new length
  1817. *
  1818. * This is identical to pskb_trim except that the caller knows that
  1819. * the skb is not cloned so we should never get an error due to out-
  1820. * of-memory.
  1821. */
  1822. static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
  1823. {
  1824. int err = pskb_trim(skb, len);
  1825. BUG_ON(err);
  1826. }
  1827. /**
  1828. * skb_orphan - orphan a buffer
  1829. * @skb: buffer to orphan
  1830. *
  1831. * If a buffer currently has an owner then we call the owner's
  1832. * destructor function and make the @skb unowned. The buffer continues
  1833. * to exist but is no longer charged to its former owner.
  1834. */
  1835. static inline void skb_orphan(struct sk_buff *skb)
  1836. {
  1837. if (skb->destructor) {
  1838. skb->destructor(skb);
  1839. skb->destructor = NULL;
  1840. skb->sk = NULL;
  1841. } else {
  1842. BUG_ON(skb->sk);
  1843. }
  1844. }
  1845. /**
  1846. * skb_orphan_frags - orphan the frags contained in a buffer
  1847. * @skb: buffer to orphan frags from
  1848. * @gfp_mask: allocation mask for replacement pages
  1849. *
  1850. * For each frag in the SKB which needs a destructor (i.e. has an
  1851. * owner) create a copy of that frag and release the original
  1852. * page by calling the destructor.
  1853. */
  1854. static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
  1855. {
  1856. if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
  1857. return 0;
  1858. return skb_copy_ubufs(skb, gfp_mask);
  1859. }
  1860. /**
  1861. * __skb_queue_purge - empty a list
  1862. * @list: list to empty
  1863. *
  1864. * Delete all buffers on an &sk_buff list. Each buffer is removed from
  1865. * the list and one reference dropped. This function does not take the
  1866. * list lock and the caller must hold the relevant locks to use it.
  1867. */
  1868. void skb_queue_purge(struct sk_buff_head *list);
  1869. static inline void __skb_queue_purge(struct sk_buff_head *list)
  1870. {
  1871. struct sk_buff *skb;
  1872. while ((skb = __skb_dequeue(list)) != NULL)
  1873. kfree_skb(skb);
  1874. }
  1875. #define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
  1876. #define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
  1877. #define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
  1878. void *netdev_alloc_frag(unsigned int fragsz);
  1879. struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
  1880. gfp_t gfp_mask);
  1881. /**
  1882. * netdev_alloc_skb - allocate an skbuff for rx on a specific device
  1883. * @dev: network device to receive on
  1884. * @length: length to allocate
  1885. *
  1886. * Allocate a new &sk_buff and assign it a usage count of one. The
  1887. * buffer has unspecified headroom built in. Users should allocate
  1888. * the headroom they think they need without accounting for the
  1889. * built in space. The built in space is used for optimisations.
  1890. *
  1891. * %NULL is returned if there is no free memory. Although this function
  1892. * allocates memory it can be called from an interrupt.
  1893. */
  1894. static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
  1895. unsigned int length)
  1896. {
  1897. return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
  1898. }
  1899. /* legacy helper around __netdev_alloc_skb() */
  1900. static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
  1901. gfp_t gfp_mask)
  1902. {
  1903. return __netdev_alloc_skb(NULL, length, gfp_mask);
  1904. }
  1905. /* legacy helper around netdev_alloc_skb() */
  1906. static inline struct sk_buff *dev_alloc_skb(unsigned int length)
  1907. {
  1908. return netdev_alloc_skb(NULL, length);
  1909. }
  1910. static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
  1911. unsigned int length, gfp_t gfp)
  1912. {
  1913. struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
  1914. if (NET_IP_ALIGN && skb)
  1915. skb_reserve(skb, NET_IP_ALIGN);
  1916. return skb;
  1917. }
  1918. static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
  1919. unsigned int length)
  1920. {
  1921. return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
  1922. }
  1923. void *napi_alloc_frag(unsigned int fragsz);
  1924. struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
  1925. unsigned int length, gfp_t gfp_mask);
  1926. static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
  1927. unsigned int length)
  1928. {
  1929. return __napi_alloc_skb(napi, length, GFP_ATOMIC);
  1930. }
  1931. /**
  1932. * __dev_alloc_pages - allocate page for network Rx
  1933. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1934. * @order: size of the allocation
  1935. *
  1936. * Allocate a new page.
  1937. *
  1938. * %NULL is returned if there is no free memory.
  1939. */
  1940. static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
  1941. unsigned int order)
  1942. {
  1943. /* This piece of code contains several assumptions.
  1944. * 1. This is for device Rx, therefor a cold page is preferred.
  1945. * 2. The expectation is the user wants a compound page.
  1946. * 3. If requesting a order 0 page it will not be compound
  1947. * due to the check to see if order has a value in prep_new_page
  1948. * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
  1949. * code in gfp_to_alloc_flags that should be enforcing this.
  1950. */
  1951. gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
  1952. return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
  1953. }
  1954. static inline struct page *dev_alloc_pages(unsigned int order)
  1955. {
  1956. return __dev_alloc_pages(GFP_ATOMIC, order);
  1957. }
  1958. /**
  1959. * __dev_alloc_page - allocate a page for network Rx
  1960. * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
  1961. *
  1962. * Allocate a new page.
  1963. *
  1964. * %NULL is returned if there is no free memory.
  1965. */
  1966. static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
  1967. {
  1968. return __dev_alloc_pages(gfp_mask, 0);
  1969. }
  1970. static inline struct page *dev_alloc_page(void)
  1971. {
  1972. return __dev_alloc_page(GFP_ATOMIC);
  1973. }
  1974. /**
  1975. * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
  1976. * @page: The page that was allocated from skb_alloc_page
  1977. * @skb: The skb that may need pfmemalloc set
  1978. */
  1979. static inline void skb_propagate_pfmemalloc(struct page *page,
  1980. struct sk_buff *skb)
  1981. {
  1982. if (page && page->pfmemalloc)
  1983. skb->pfmemalloc = true;
  1984. }
  1985. /**
  1986. * skb_frag_page - retrieve the page referred to by a paged fragment
  1987. * @frag: the paged fragment
  1988. *
  1989. * Returns the &struct page associated with @frag.
  1990. */
  1991. static inline struct page *skb_frag_page(const skb_frag_t *frag)
  1992. {
  1993. return frag->page.p;
  1994. }
  1995. /**
  1996. * __skb_frag_ref - take an addition reference on a paged fragment.
  1997. * @frag: the paged fragment
  1998. *
  1999. * Takes an additional reference on the paged fragment @frag.
  2000. */
  2001. static inline void __skb_frag_ref(skb_frag_t *frag)
  2002. {
  2003. get_page(skb_frag_page(frag));
  2004. }
  2005. /**
  2006. * skb_frag_ref - take an addition reference on a paged fragment of an skb.
  2007. * @skb: the buffer
  2008. * @f: the fragment offset.
  2009. *
  2010. * Takes an additional reference on the @f'th paged fragment of @skb.
  2011. */
  2012. static inline void skb_frag_ref(struct sk_buff *skb, int f)
  2013. {
  2014. __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
  2015. }
  2016. /**
  2017. * __skb_frag_unref - release a reference on a paged fragment.
  2018. * @frag: the paged fragment
  2019. *
  2020. * Releases a reference on the paged fragment @frag.
  2021. */
  2022. static inline void __skb_frag_unref(skb_frag_t *frag)
  2023. {
  2024. put_page(skb_frag_page(frag));
  2025. }
  2026. /**
  2027. * skb_frag_unref - release a reference on a paged fragment of an skb.
  2028. * @skb: the buffer
  2029. * @f: the fragment offset
  2030. *
  2031. * Releases a reference on the @f'th paged fragment of @skb.
  2032. */
  2033. static inline void skb_frag_unref(struct sk_buff *skb, int f)
  2034. {
  2035. __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
  2036. }
  2037. /**
  2038. * skb_frag_address - gets the address of the data contained in a paged fragment
  2039. * @frag: the paged fragment buffer
  2040. *
  2041. * Returns the address of the data within @frag. The page must already
  2042. * be mapped.
  2043. */
  2044. static inline void *skb_frag_address(const skb_frag_t *frag)
  2045. {
  2046. return page_address(skb_frag_page(frag)) + frag->page_offset;
  2047. }
  2048. /**
  2049. * skb_frag_address_safe - gets the address of the data contained in a paged fragment
  2050. * @frag: the paged fragment buffer
  2051. *
  2052. * Returns the address of the data within @frag. Checks that the page
  2053. * is mapped and returns %NULL otherwise.
  2054. */
  2055. static inline void *skb_frag_address_safe(const skb_frag_t *frag)
  2056. {
  2057. void *ptr = page_address(skb_frag_page(frag));
  2058. if (unlikely(!ptr))
  2059. return NULL;
  2060. return ptr + frag->page_offset;
  2061. }
  2062. /**
  2063. * __skb_frag_set_page - sets the page contained in a paged fragment
  2064. * @frag: the paged fragment
  2065. * @page: the page to set
  2066. *
  2067. * Sets the fragment @frag to contain @page.
  2068. */
  2069. static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
  2070. {
  2071. frag->page.p = page;
  2072. }
  2073. /**
  2074. * skb_frag_set_page - sets the page contained in a paged fragment of an skb
  2075. * @skb: the buffer
  2076. * @f: the fragment offset
  2077. * @page: the page to set
  2078. *
  2079. * Sets the @f'th fragment of @skb to contain @page.
  2080. */
  2081. static inline void skb_frag_set_page(struct sk_buff *skb, int f,
  2082. struct page *page)
  2083. {
  2084. __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
  2085. }
  2086. bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
  2087. /**
  2088. * skb_frag_dma_map - maps a paged fragment via the DMA API
  2089. * @dev: the device to map the fragment to
  2090. * @frag: the paged fragment to map
  2091. * @offset: the offset within the fragment (starting at the
  2092. * fragment's own offset)
  2093. * @size: the number of bytes to map
  2094. * @dir: the direction of the mapping (%PCI_DMA_*)
  2095. *
  2096. * Maps the page associated with @frag to @device.
  2097. */
  2098. static inline dma_addr_t skb_frag_dma_map(struct device *dev,
  2099. const skb_frag_t *frag,
  2100. size_t offset, size_t size,
  2101. enum dma_data_direction dir)
  2102. {
  2103. return dma_map_page(dev, skb_frag_page(frag),
  2104. frag->page_offset + offset, size, dir);
  2105. }
  2106. static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
  2107. gfp_t gfp_mask)
  2108. {
  2109. return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
  2110. }
  2111. static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
  2112. gfp_t gfp_mask)
  2113. {
  2114. return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
  2115. }
  2116. /**
  2117. * skb_clone_writable - is the header of a clone writable
  2118. * @skb: buffer to check
  2119. * @len: length up to which to write
  2120. *
  2121. * Returns true if modifying the header part of the cloned buffer
  2122. * does not requires the data to be copied.
  2123. */
  2124. static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
  2125. {
  2126. return !skb_header_cloned(skb) &&
  2127. skb_headroom(skb) + len <= skb->hdr_len;
  2128. }
  2129. static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
  2130. int cloned)
  2131. {
  2132. int delta = 0;
  2133. if (headroom > skb_headroom(skb))
  2134. delta = headroom - skb_headroom(skb);
  2135. if (delta || cloned)
  2136. return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
  2137. GFP_ATOMIC);
  2138. return 0;
  2139. }
  2140. /**
  2141. * skb_cow - copy header of skb when it is required
  2142. * @skb: buffer to cow
  2143. * @headroom: needed headroom
  2144. *
  2145. * If the skb passed lacks sufficient headroom or its data part
  2146. * is shared, data is reallocated. If reallocation fails, an error
  2147. * is returned and original skb is not changed.
  2148. *
  2149. * The result is skb with writable area skb->head...skb->tail
  2150. * and at least @headroom of space at head.
  2151. */
  2152. static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
  2153. {
  2154. return __skb_cow(skb, headroom, skb_cloned(skb));
  2155. }
  2156. /**
  2157. * skb_cow_head - skb_cow but only making the head writable
  2158. * @skb: buffer to cow
  2159. * @headroom: needed headroom
  2160. *
  2161. * This function is identical to skb_cow except that we replace the
  2162. * skb_cloned check by skb_header_cloned. It should be used when
  2163. * you only need to push on some header and do not need to modify
  2164. * the data.
  2165. */
  2166. static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
  2167. {
  2168. return __skb_cow(skb, headroom, skb_header_cloned(skb));
  2169. }
  2170. /**
  2171. * skb_padto - pad an skbuff up to a minimal size
  2172. * @skb: buffer to pad
  2173. * @len: minimal length
  2174. *
  2175. * Pads up a buffer to ensure the trailing bytes exist and are
  2176. * blanked. If the buffer already contains sufficient data it
  2177. * is untouched. Otherwise it is extended. Returns zero on
  2178. * success. The skb is freed on error.
  2179. */
  2180. static inline int skb_padto(struct sk_buff *skb, unsigned int len)
  2181. {
  2182. unsigned int size = skb->len;
  2183. if (likely(size >= len))
  2184. return 0;
  2185. return skb_pad(skb, len - size);
  2186. }
  2187. /**
  2188. * skb_put_padto - increase size and pad an skbuff up to a minimal size
  2189. * @skb: buffer to pad
  2190. * @len: minimal length
  2191. *
  2192. * Pads up a buffer to ensure the trailing bytes exist and are
  2193. * blanked. If the buffer already contains sufficient data it
  2194. * is untouched. Otherwise it is extended. Returns zero on
  2195. * success. The skb is freed on error.
  2196. */
  2197. static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
  2198. {
  2199. unsigned int size = skb->len;
  2200. if (unlikely(size < len)) {
  2201. len -= size;
  2202. if (skb_pad(skb, len))
  2203. return -ENOMEM;
  2204. __skb_put(skb, len);
  2205. }
  2206. return 0;
  2207. }
  2208. static inline int skb_add_data(struct sk_buff *skb,
  2209. struct iov_iter *from, int copy)
  2210. {
  2211. const int off = skb->len;
  2212. if (skb->ip_summed == CHECKSUM_NONE) {
  2213. __wsum csum = 0;
  2214. if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
  2215. &csum, from) == copy) {
  2216. skb->csum = csum_block_add(skb->csum, csum, off);
  2217. return 0;
  2218. }
  2219. } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
  2220. return 0;
  2221. __skb_trim(skb, off);
  2222. return -EFAULT;
  2223. }
  2224. static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
  2225. const struct page *page, int off)
  2226. {
  2227. if (i) {
  2228. const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
  2229. return page == skb_frag_page(frag) &&
  2230. off == frag->page_offset + skb_frag_size(frag);
  2231. }
  2232. return false;
  2233. }
  2234. static inline int __skb_linearize(struct sk_buff *skb)
  2235. {
  2236. return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
  2237. }
  2238. /**
  2239. * skb_linearize - convert paged skb to linear one
  2240. * @skb: buffer to linarize
  2241. *
  2242. * If there is no free memory -ENOMEM is returned, otherwise zero
  2243. * is returned and the old skb data released.
  2244. */
  2245. static inline int skb_linearize(struct sk_buff *skb)
  2246. {
  2247. return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
  2248. }
  2249. /**
  2250. * skb_has_shared_frag - can any frag be overwritten
  2251. * @skb: buffer to test
  2252. *
  2253. * Return true if the skb has at least one frag that might be modified
  2254. * by an external entity (as in vmsplice()/sendfile())
  2255. */
  2256. static inline bool skb_has_shared_frag(const struct sk_buff *skb)
  2257. {
  2258. return skb_is_nonlinear(skb) &&
  2259. skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
  2260. }
  2261. /**
  2262. * skb_linearize_cow - make sure skb is linear and writable
  2263. * @skb: buffer to process
  2264. *
  2265. * If there is no free memory -ENOMEM is returned, otherwise zero
  2266. * is returned and the old skb data released.
  2267. */
  2268. static inline int skb_linearize_cow(struct sk_buff *skb)
  2269. {
  2270. return skb_is_nonlinear(skb) || skb_cloned(skb) ?
  2271. __skb_linearize(skb) : 0;
  2272. }
  2273. /**
  2274. * skb_postpull_rcsum - update checksum for received skb after pull
  2275. * @skb: buffer to update
  2276. * @start: start of data before pull
  2277. * @len: length of data pulled
  2278. *
  2279. * After doing a pull on a received packet, you need to call this to
  2280. * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
  2281. * CHECKSUM_NONE so that it can be recomputed from scratch.
  2282. */
  2283. static inline void skb_postpull_rcsum(struct sk_buff *skb,
  2284. const void *start, unsigned int len)
  2285. {
  2286. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2287. skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
  2288. }
  2289. unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
  2290. /**
  2291. * pskb_trim_rcsum - trim received skb and update checksum
  2292. * @skb: buffer to trim
  2293. * @len: new length
  2294. *
  2295. * This is exactly the same as pskb_trim except that it ensures the
  2296. * checksum of received packets are still valid after the operation.
  2297. */
  2298. static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
  2299. {
  2300. if (likely(len >= skb->len))
  2301. return 0;
  2302. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2303. skb->ip_summed = CHECKSUM_NONE;
  2304. return __pskb_trim(skb, len);
  2305. }
  2306. #define skb_queue_walk(queue, skb) \
  2307. for (skb = (queue)->next; \
  2308. skb != (struct sk_buff *)(queue); \
  2309. skb = skb->next)
  2310. #define skb_queue_walk_safe(queue, skb, tmp) \
  2311. for (skb = (queue)->next, tmp = skb->next; \
  2312. skb != (struct sk_buff *)(queue); \
  2313. skb = tmp, tmp = skb->next)
  2314. #define skb_queue_walk_from(queue, skb) \
  2315. for (; skb != (struct sk_buff *)(queue); \
  2316. skb = skb->next)
  2317. #define skb_queue_walk_from_safe(queue, skb, tmp) \
  2318. for (tmp = skb->next; \
  2319. skb != (struct sk_buff *)(queue); \
  2320. skb = tmp, tmp = skb->next)
  2321. #define skb_queue_reverse_walk(queue, skb) \
  2322. for (skb = (queue)->prev; \
  2323. skb != (struct sk_buff *)(queue); \
  2324. skb = skb->prev)
  2325. #define skb_queue_reverse_walk_safe(queue, skb, tmp) \
  2326. for (skb = (queue)->prev, tmp = skb->prev; \
  2327. skb != (struct sk_buff *)(queue); \
  2328. skb = tmp, tmp = skb->prev)
  2329. #define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
  2330. for (tmp = skb->prev; \
  2331. skb != (struct sk_buff *)(queue); \
  2332. skb = tmp, tmp = skb->prev)
  2333. static inline bool skb_has_frag_list(const struct sk_buff *skb)
  2334. {
  2335. return skb_shinfo(skb)->frag_list != NULL;
  2336. }
  2337. static inline void skb_frag_list_init(struct sk_buff *skb)
  2338. {
  2339. skb_shinfo(skb)->frag_list = NULL;
  2340. }
  2341. static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
  2342. {
  2343. frag->next = skb_shinfo(skb)->frag_list;
  2344. skb_shinfo(skb)->frag_list = frag;
  2345. }
  2346. #define skb_walk_frags(skb, iter) \
  2347. for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
  2348. struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
  2349. int *peeked, int *off, int *err);
  2350. struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
  2351. int *err);
  2352. unsigned int datagram_poll(struct file *file, struct socket *sock,
  2353. struct poll_table_struct *wait);
  2354. int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
  2355. struct iov_iter *to, int size);
  2356. static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
  2357. struct msghdr *msg, int size)
  2358. {
  2359. return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
  2360. }
  2361. int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
  2362. struct msghdr *msg);
  2363. int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
  2364. struct iov_iter *from, int len);
  2365. int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
  2366. void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
  2367. void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
  2368. int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
  2369. int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
  2370. int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
  2371. __wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
  2372. int len, __wsum csum);
  2373. int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
  2374. struct pipe_inode_info *pipe, unsigned int len,
  2375. unsigned int flags);
  2376. void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
  2377. unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
  2378. int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
  2379. int len, int hlen);
  2380. void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
  2381. int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
  2382. void skb_scrub_packet(struct sk_buff *skb, bool xnet);
  2383. unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
  2384. struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
  2385. struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
  2386. int skb_ensure_writable(struct sk_buff *skb, int write_len);
  2387. int skb_vlan_pop(struct sk_buff *skb);
  2388. int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
  2389. static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
  2390. {
  2391. return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2392. }
  2393. static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
  2394. {
  2395. return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
  2396. }
  2397. struct skb_checksum_ops {
  2398. __wsum (*update)(const void *mem, int len, __wsum wsum);
  2399. __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
  2400. };
  2401. __wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
  2402. __wsum csum, const struct skb_checksum_ops *ops);
  2403. __wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
  2404. __wsum csum);
  2405. static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
  2406. int len, void *data, int hlen, void *buffer)
  2407. {
  2408. if (hlen - offset >= len)
  2409. return data + offset;
  2410. if (!skb ||
  2411. skb_copy_bits(skb, offset, buffer, len) < 0)
  2412. return NULL;
  2413. return buffer;
  2414. }
  2415. static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
  2416. int len, void *buffer)
  2417. {
  2418. return __skb_header_pointer(skb, offset, len, skb->data,
  2419. skb_headlen(skb), buffer);
  2420. }
  2421. /**
  2422. * skb_needs_linearize - check if we need to linearize a given skb
  2423. * depending on the given device features.
  2424. * @skb: socket buffer to check
  2425. * @features: net device features
  2426. *
  2427. * Returns true if either:
  2428. * 1. skb has frag_list and the device doesn't support FRAGLIST, or
  2429. * 2. skb is fragmented and the device does not support SG.
  2430. */
  2431. static inline bool skb_needs_linearize(struct sk_buff *skb,
  2432. netdev_features_t features)
  2433. {
  2434. return skb_is_nonlinear(skb) &&
  2435. ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
  2436. (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
  2437. }
  2438. static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
  2439. void *to,
  2440. const unsigned int len)
  2441. {
  2442. memcpy(to, skb->data, len);
  2443. }
  2444. static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
  2445. const int offset, void *to,
  2446. const unsigned int len)
  2447. {
  2448. memcpy(to, skb->data + offset, len);
  2449. }
  2450. static inline void skb_copy_to_linear_data(struct sk_buff *skb,
  2451. const void *from,
  2452. const unsigned int len)
  2453. {
  2454. memcpy(skb->data, from, len);
  2455. }
  2456. static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
  2457. const int offset,
  2458. const void *from,
  2459. const unsigned int len)
  2460. {
  2461. memcpy(skb->data + offset, from, len);
  2462. }
  2463. void skb_init(void);
  2464. static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
  2465. {
  2466. return skb->tstamp;
  2467. }
  2468. /**
  2469. * skb_get_timestamp - get timestamp from a skb
  2470. * @skb: skb to get stamp from
  2471. * @stamp: pointer to struct timeval to store stamp in
  2472. *
  2473. * Timestamps are stored in the skb as offsets to a base timestamp.
  2474. * This function converts the offset back to a struct timeval and stores
  2475. * it in stamp.
  2476. */
  2477. static inline void skb_get_timestamp(const struct sk_buff *skb,
  2478. struct timeval *stamp)
  2479. {
  2480. *stamp = ktime_to_timeval(skb->tstamp);
  2481. }
  2482. static inline void skb_get_timestampns(const struct sk_buff *skb,
  2483. struct timespec *stamp)
  2484. {
  2485. *stamp = ktime_to_timespec(skb->tstamp);
  2486. }
  2487. static inline void __net_timestamp(struct sk_buff *skb)
  2488. {
  2489. skb->tstamp = ktime_get_real();
  2490. }
  2491. static inline ktime_t net_timedelta(ktime_t t)
  2492. {
  2493. return ktime_sub(ktime_get_real(), t);
  2494. }
  2495. static inline ktime_t net_invalid_timestamp(void)
  2496. {
  2497. return ktime_set(0, 0);
  2498. }
  2499. struct sk_buff *skb_clone_sk(struct sk_buff *skb);
  2500. #ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
  2501. void skb_clone_tx_timestamp(struct sk_buff *skb);
  2502. bool skb_defer_rx_timestamp(struct sk_buff *skb);
  2503. #else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
  2504. static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
  2505. {
  2506. }
  2507. static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
  2508. {
  2509. return false;
  2510. }
  2511. #endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
  2512. /**
  2513. * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
  2514. *
  2515. * PHY drivers may accept clones of transmitted packets for
  2516. * timestamping via their phy_driver.txtstamp method. These drivers
  2517. * must call this function to return the skb back to the stack, with
  2518. * or without a timestamp.
  2519. *
  2520. * @skb: clone of the the original outgoing packet
  2521. * @hwtstamps: hardware time stamps, may be NULL if not available
  2522. *
  2523. */
  2524. void skb_complete_tx_timestamp(struct sk_buff *skb,
  2525. struct skb_shared_hwtstamps *hwtstamps);
  2526. void __skb_tstamp_tx(struct sk_buff *orig_skb,
  2527. struct skb_shared_hwtstamps *hwtstamps,
  2528. struct sock *sk, int tstype);
  2529. /**
  2530. * skb_tstamp_tx - queue clone of skb with send time stamps
  2531. * @orig_skb: the original outgoing packet
  2532. * @hwtstamps: hardware time stamps, may be NULL if not available
  2533. *
  2534. * If the skb has a socket associated, then this function clones the
  2535. * skb (thus sharing the actual data and optional structures), stores
  2536. * the optional hardware time stamping information (if non NULL) or
  2537. * generates a software time stamp (otherwise), then queues the clone
  2538. * to the error queue of the socket. Errors are silently ignored.
  2539. */
  2540. void skb_tstamp_tx(struct sk_buff *orig_skb,
  2541. struct skb_shared_hwtstamps *hwtstamps);
  2542. static inline void sw_tx_timestamp(struct sk_buff *skb)
  2543. {
  2544. if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
  2545. !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
  2546. skb_tstamp_tx(skb, NULL);
  2547. }
  2548. /**
  2549. * skb_tx_timestamp() - Driver hook for transmit timestamping
  2550. *
  2551. * Ethernet MAC Drivers should call this function in their hard_xmit()
  2552. * function immediately before giving the sk_buff to the MAC hardware.
  2553. *
  2554. * Specifically, one should make absolutely sure that this function is
  2555. * called before TX completion of this packet can trigger. Otherwise
  2556. * the packet could potentially already be freed.
  2557. *
  2558. * @skb: A socket buffer.
  2559. */
  2560. static inline void skb_tx_timestamp(struct sk_buff *skb)
  2561. {
  2562. skb_clone_tx_timestamp(skb);
  2563. sw_tx_timestamp(skb);
  2564. }
  2565. /**
  2566. * skb_complete_wifi_ack - deliver skb with wifi status
  2567. *
  2568. * @skb: the original outgoing packet
  2569. * @acked: ack status
  2570. *
  2571. */
  2572. void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
  2573. __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
  2574. __sum16 __skb_checksum_complete(struct sk_buff *skb);
  2575. static inline int skb_csum_unnecessary(const struct sk_buff *skb)
  2576. {
  2577. return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
  2578. skb->csum_valid ||
  2579. (skb->ip_summed == CHECKSUM_PARTIAL &&
  2580. skb_checksum_start_offset(skb) >= 0));
  2581. }
  2582. /**
  2583. * skb_checksum_complete - Calculate checksum of an entire packet
  2584. * @skb: packet to process
  2585. *
  2586. * This function calculates the checksum over the entire packet plus
  2587. * the value of skb->csum. The latter can be used to supply the
  2588. * checksum of a pseudo header as used by TCP/UDP. It returns the
  2589. * checksum.
  2590. *
  2591. * For protocols that contain complete checksums such as ICMP/TCP/UDP,
  2592. * this function can be used to verify that checksum on received
  2593. * packets. In that case the function should return zero if the
  2594. * checksum is correct. In particular, this function will return zero
  2595. * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
  2596. * hardware has already verified the correctness of the checksum.
  2597. */
  2598. static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
  2599. {
  2600. return skb_csum_unnecessary(skb) ?
  2601. 0 : __skb_checksum_complete(skb);
  2602. }
  2603. static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
  2604. {
  2605. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2606. if (skb->csum_level == 0)
  2607. skb->ip_summed = CHECKSUM_NONE;
  2608. else
  2609. skb->csum_level--;
  2610. }
  2611. }
  2612. static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
  2613. {
  2614. if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
  2615. if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
  2616. skb->csum_level++;
  2617. } else if (skb->ip_summed == CHECKSUM_NONE) {
  2618. skb->ip_summed = CHECKSUM_UNNECESSARY;
  2619. skb->csum_level = 0;
  2620. }
  2621. }
  2622. static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
  2623. {
  2624. /* Mark current checksum as bad (typically called from GRO
  2625. * path). In the case that ip_summed is CHECKSUM_NONE
  2626. * this must be the first checksum encountered in the packet.
  2627. * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
  2628. * checksum after the last one validated. For UDP, a zero
  2629. * checksum can not be marked as bad.
  2630. */
  2631. if (skb->ip_summed == CHECKSUM_NONE ||
  2632. skb->ip_summed == CHECKSUM_UNNECESSARY)
  2633. skb->csum_bad = 1;
  2634. }
  2635. /* Check if we need to perform checksum complete validation.
  2636. *
  2637. * Returns true if checksum complete is needed, false otherwise
  2638. * (either checksum is unnecessary or zero checksum is allowed).
  2639. */
  2640. static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
  2641. bool zero_okay,
  2642. __sum16 check)
  2643. {
  2644. if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
  2645. skb->csum_valid = 1;
  2646. __skb_decr_checksum_unnecessary(skb);
  2647. return false;
  2648. }
  2649. return true;
  2650. }
  2651. /* For small packets <= CHECKSUM_BREAK peform checksum complete directly
  2652. * in checksum_init.
  2653. */
  2654. #define CHECKSUM_BREAK 76
  2655. /* Unset checksum-complete
  2656. *
  2657. * Unset checksum complete can be done when packet is being modified
  2658. * (uncompressed for instance) and checksum-complete value is
  2659. * invalidated.
  2660. */
  2661. static inline void skb_checksum_complete_unset(struct sk_buff *skb)
  2662. {
  2663. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2664. skb->ip_summed = CHECKSUM_NONE;
  2665. }
  2666. /* Validate (init) checksum based on checksum complete.
  2667. *
  2668. * Return values:
  2669. * 0: checksum is validated or try to in skb_checksum_complete. In the latter
  2670. * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
  2671. * checksum is stored in skb->csum for use in __skb_checksum_complete
  2672. * non-zero: value of invalid checksum
  2673. *
  2674. */
  2675. static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
  2676. bool complete,
  2677. __wsum psum)
  2678. {
  2679. if (skb->ip_summed == CHECKSUM_COMPLETE) {
  2680. if (!csum_fold(csum_add(psum, skb->csum))) {
  2681. skb->csum_valid = 1;
  2682. return 0;
  2683. }
  2684. } else if (skb->csum_bad) {
  2685. /* ip_summed == CHECKSUM_NONE in this case */
  2686. return 1;
  2687. }
  2688. skb->csum = psum;
  2689. if (complete || skb->len <= CHECKSUM_BREAK) {
  2690. __sum16 csum;
  2691. csum = __skb_checksum_complete(skb);
  2692. skb->csum_valid = !csum;
  2693. return csum;
  2694. }
  2695. return 0;
  2696. }
  2697. static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
  2698. {
  2699. return 0;
  2700. }
  2701. /* Perform checksum validate (init). Note that this is a macro since we only
  2702. * want to calculate the pseudo header which is an input function if necessary.
  2703. * First we try to validate without any computation (checksum unnecessary) and
  2704. * then calculate based on checksum complete calling the function to compute
  2705. * pseudo header.
  2706. *
  2707. * Return values:
  2708. * 0: checksum is validated or try to in skb_checksum_complete
  2709. * non-zero: value of invalid checksum
  2710. */
  2711. #define __skb_checksum_validate(skb, proto, complete, \
  2712. zero_okay, check, compute_pseudo) \
  2713. ({ \
  2714. __sum16 __ret = 0; \
  2715. skb->csum_valid = 0; \
  2716. if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
  2717. __ret = __skb_checksum_validate_complete(skb, \
  2718. complete, compute_pseudo(skb, proto)); \
  2719. __ret; \
  2720. })
  2721. #define skb_checksum_init(skb, proto, compute_pseudo) \
  2722. __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
  2723. #define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
  2724. __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
  2725. #define skb_checksum_validate(skb, proto, compute_pseudo) \
  2726. __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
  2727. #define skb_checksum_validate_zero_check(skb, proto, check, \
  2728. compute_pseudo) \
  2729. __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
  2730. #define skb_checksum_simple_validate(skb) \
  2731. __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
  2732. static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
  2733. {
  2734. return (skb->ip_summed == CHECKSUM_NONE &&
  2735. skb->csum_valid && !skb->csum_bad);
  2736. }
  2737. static inline void __skb_checksum_convert(struct sk_buff *skb,
  2738. __sum16 check, __wsum pseudo)
  2739. {
  2740. skb->csum = ~pseudo;
  2741. skb->ip_summed = CHECKSUM_COMPLETE;
  2742. }
  2743. #define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
  2744. do { \
  2745. if (__skb_checksum_convert_check(skb)) \
  2746. __skb_checksum_convert(skb, check, \
  2747. compute_pseudo(skb, proto)); \
  2748. } while (0)
  2749. static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
  2750. u16 start, u16 offset)
  2751. {
  2752. skb->ip_summed = CHECKSUM_PARTIAL;
  2753. skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
  2754. skb->csum_offset = offset - start;
  2755. }
  2756. /* Update skbuf and packet to reflect the remote checksum offload operation.
  2757. * When called, ptr indicates the starting point for skb->csum when
  2758. * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
  2759. * here, skb_postpull_rcsum is done so skb->csum start is ptr.
  2760. */
  2761. static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
  2762. int start, int offset, bool nopartial)
  2763. {
  2764. __wsum delta;
  2765. if (!nopartial) {
  2766. skb_remcsum_adjust_partial(skb, ptr, start, offset);
  2767. return;
  2768. }
  2769. if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
  2770. __skb_checksum_complete(skb);
  2771. skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
  2772. }
  2773. delta = remcsum_adjust(ptr, skb->csum, start, offset);
  2774. /* Adjust skb->csum since we changed the packet */
  2775. skb->csum = csum_add(skb->csum, delta);
  2776. }
  2777. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2778. void nf_conntrack_destroy(struct nf_conntrack *nfct);
  2779. static inline void nf_conntrack_put(struct nf_conntrack *nfct)
  2780. {
  2781. if (nfct && atomic_dec_and_test(&nfct->use))
  2782. nf_conntrack_destroy(nfct);
  2783. }
  2784. static inline void nf_conntrack_get(struct nf_conntrack *nfct)
  2785. {
  2786. if (nfct)
  2787. atomic_inc(&nfct->use);
  2788. }
  2789. #endif
  2790. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2791. static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
  2792. {
  2793. if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
  2794. kfree(nf_bridge);
  2795. }
  2796. static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
  2797. {
  2798. if (nf_bridge)
  2799. atomic_inc(&nf_bridge->use);
  2800. }
  2801. #endif /* CONFIG_BRIDGE_NETFILTER */
  2802. static inline void nf_reset(struct sk_buff *skb)
  2803. {
  2804. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2805. nf_conntrack_put(skb->nfct);
  2806. skb->nfct = NULL;
  2807. #endif
  2808. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2809. nf_bridge_put(skb->nf_bridge);
  2810. skb->nf_bridge = NULL;
  2811. #endif
  2812. }
  2813. static inline void nf_reset_trace(struct sk_buff *skb)
  2814. {
  2815. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2816. skb->nf_trace = 0;
  2817. #endif
  2818. }
  2819. /* Note: This doesn't put any conntrack and bridge info in dst. */
  2820. static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
  2821. bool copy)
  2822. {
  2823. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2824. dst->nfct = src->nfct;
  2825. nf_conntrack_get(src->nfct);
  2826. if (copy)
  2827. dst->nfctinfo = src->nfctinfo;
  2828. #endif
  2829. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2830. dst->nf_bridge = src->nf_bridge;
  2831. nf_bridge_get(src->nf_bridge);
  2832. #endif
  2833. #if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
  2834. if (copy)
  2835. dst->nf_trace = src->nf_trace;
  2836. #endif
  2837. }
  2838. static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
  2839. {
  2840. #if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
  2841. nf_conntrack_put(dst->nfct);
  2842. #endif
  2843. #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
  2844. nf_bridge_put(dst->nf_bridge);
  2845. #endif
  2846. __nf_copy(dst, src, true);
  2847. }
  2848. #ifdef CONFIG_NETWORK_SECMARK
  2849. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2850. {
  2851. to->secmark = from->secmark;
  2852. }
  2853. static inline void skb_init_secmark(struct sk_buff *skb)
  2854. {
  2855. skb->secmark = 0;
  2856. }
  2857. #else
  2858. static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
  2859. { }
  2860. static inline void skb_init_secmark(struct sk_buff *skb)
  2861. { }
  2862. #endif
  2863. static inline bool skb_irq_freeable(const struct sk_buff *skb)
  2864. {
  2865. return !skb->destructor &&
  2866. #if IS_ENABLED(CONFIG_XFRM)
  2867. !skb->sp &&
  2868. #endif
  2869. #if IS_ENABLED(CONFIG_NF_CONNTRACK)
  2870. !skb->nfct &&
  2871. #endif
  2872. !skb->_skb_refdst &&
  2873. !skb_has_frag_list(skb);
  2874. }
  2875. static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
  2876. {
  2877. skb->queue_mapping = queue_mapping;
  2878. }
  2879. static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
  2880. {
  2881. return skb->queue_mapping;
  2882. }
  2883. static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
  2884. {
  2885. to->queue_mapping = from->queue_mapping;
  2886. }
  2887. static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
  2888. {
  2889. skb->queue_mapping = rx_queue + 1;
  2890. }
  2891. static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
  2892. {
  2893. return skb->queue_mapping - 1;
  2894. }
  2895. static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
  2896. {
  2897. return skb->queue_mapping != 0;
  2898. }
  2899. u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
  2900. unsigned int num_tx_queues);
  2901. static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
  2902. {
  2903. #ifdef CONFIG_XFRM
  2904. return skb->sp;
  2905. #else
  2906. return NULL;
  2907. #endif
  2908. }
  2909. /* Keeps track of mac header offset relative to skb->head.
  2910. * It is useful for TSO of Tunneling protocol. e.g. GRE.
  2911. * For non-tunnel skb it points to skb_mac_header() and for
  2912. * tunnel skb it points to outer mac header.
  2913. * Keeps track of level of encapsulation of network headers.
  2914. */
  2915. struct skb_gso_cb {
  2916. int mac_offset;
  2917. int encap_level;
  2918. __u16 csum_start;
  2919. };
  2920. #define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
  2921. static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
  2922. {
  2923. return (skb_mac_header(inner_skb) - inner_skb->head) -
  2924. SKB_GSO_CB(inner_skb)->mac_offset;
  2925. }
  2926. static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
  2927. {
  2928. int new_headroom, headroom;
  2929. int ret;
  2930. headroom = skb_headroom(skb);
  2931. ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
  2932. if (ret)
  2933. return ret;
  2934. new_headroom = skb_headroom(skb);
  2935. SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
  2936. return 0;
  2937. }
  2938. /* Compute the checksum for a gso segment. First compute the checksum value
  2939. * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
  2940. * then add in skb->csum (checksum from csum_start to end of packet).
  2941. * skb->csum and csum_start are then updated to reflect the checksum of the
  2942. * resultant packet starting from the transport header-- the resultant checksum
  2943. * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
  2944. * header.
  2945. */
  2946. static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
  2947. {
  2948. int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
  2949. skb_transport_offset(skb);
  2950. __u16 csum;
  2951. csum = csum_fold(csum_partial(skb_transport_header(skb),
  2952. plen, skb->csum));
  2953. skb->csum = res;
  2954. SKB_GSO_CB(skb)->csum_start -= plen;
  2955. return csum;
  2956. }
  2957. static inline bool skb_is_gso(const struct sk_buff *skb)
  2958. {
  2959. return skb_shinfo(skb)->gso_size;
  2960. }
  2961. /* Note: Should be called only if skb_is_gso(skb) is true */
  2962. static inline bool skb_is_gso_v6(const struct sk_buff *skb)
  2963. {
  2964. return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
  2965. }
  2966. void __skb_warn_lro_forwarding(const struct sk_buff *skb);
  2967. static inline bool skb_warn_if_lro(const struct sk_buff *skb)
  2968. {
  2969. /* LRO sets gso_size but not gso_type, whereas if GSO is really
  2970. * wanted then gso_type will be set. */
  2971. const struct skb_shared_info *shinfo = skb_shinfo(skb);
  2972. if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
  2973. unlikely(shinfo->gso_type == 0)) {
  2974. __skb_warn_lro_forwarding(skb);
  2975. return true;
  2976. }
  2977. return false;
  2978. }
  2979. static inline void skb_forward_csum(struct sk_buff *skb)
  2980. {
  2981. /* Unfortunately we don't support this one. Any brave souls? */
  2982. if (skb->ip_summed == CHECKSUM_COMPLETE)
  2983. skb->ip_summed = CHECKSUM_NONE;
  2984. }
  2985. /**
  2986. * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
  2987. * @skb: skb to check
  2988. *
  2989. * fresh skbs have their ip_summed set to CHECKSUM_NONE.
  2990. * Instead of forcing ip_summed to CHECKSUM_NONE, we can
  2991. * use this helper, to document places where we make this assertion.
  2992. */
  2993. static inline void skb_checksum_none_assert(const struct sk_buff *skb)
  2994. {
  2995. #ifdef DEBUG
  2996. BUG_ON(skb->ip_summed != CHECKSUM_NONE);
  2997. #endif
  2998. }
  2999. bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
  3000. int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
  3001. u32 skb_get_poff(const struct sk_buff *skb);
  3002. u32 __skb_get_poff(const struct sk_buff *skb, void *data,
  3003. const struct flow_keys *keys, int hlen);
  3004. /**
  3005. * skb_head_is_locked - Determine if the skb->head is locked down
  3006. * @skb: skb to check
  3007. *
  3008. * The head on skbs build around a head frag can be removed if they are
  3009. * not cloned. This function returns true if the skb head is locked down
  3010. * due to either being allocated via kmalloc, or by being a clone with
  3011. * multiple references to the head.
  3012. */
  3013. static inline bool skb_head_is_locked(const struct sk_buff *skb)
  3014. {
  3015. return !skb->head_frag || skb_cloned(skb);
  3016. }
  3017. /**
  3018. * skb_gso_network_seglen - Return length of individual segments of a gso packet
  3019. *
  3020. * @skb: GSO skb
  3021. *
  3022. * skb_gso_network_seglen is used to determine the real size of the
  3023. * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
  3024. *
  3025. * The MAC/L2 header is not accounted for.
  3026. */
  3027. static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
  3028. {
  3029. unsigned int hdr_len = skb_transport_header(skb) -
  3030. skb_network_header(skb);
  3031. return hdr_len + skb_gso_transport_seglen(skb);
  3032. }
  3033. #endif /* __KERNEL__ */
  3034. #endif /* _LINUX_SKBUFF_H */